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

2 Seals and Salmon Farms

2.1 Introduction

The most common application for marine acoustic deterrent devices in Scotland is the aquaculture industry. Furthermore, much of the relevant research on the effects of acoustic deterrent devices has been conducted in the context of marine aquaculture. This is a reflection of the economic losses sustained by aquaculture as a result of seal depredation of caged fish. Much of this report is therefore focused on acoustic deterrents used on fish farms, and other methods that might be useful in minimising seal damage.

2.1.1 Size and Nature of the Problem

Salmon aquaculture is one of Scotland's most important rural industries, producing more than 158,000 tonnes of salmon at 254 sites in 2011, with a farm gate value of more than £400 million. The industry directly employed 1,064 people in 2010, with a further 850 in directly related onshore jobs and an estimated 4,500 jobs in downstream processing (Marine Scotland Science, 2011). The industry now accounts for about half of all Scottish food exports by value. Scotland's National Marine Plan pre-consultation draft (2011) aims for sustainable production of marine finfish to increase to 210,000 tonnes by 2020. Predator control is highlighted as a potential environmental impact within this plan, which recognises a need for high standards of environmental protection at every stage of fish farm planning, operation and regulation. The Scottish Government's renewed strategic framework for Scottish aquaculture, 'A Fresh Start' (Marine Scotland, 2009), also highlights the need for the growth of the aquaculture industry to be sustainable, within the carrying capacity of the environment and balanced against the needs of others.

Hawkins (1985) was the first to review the problem of seal depredation in Scotland, and found predation by seals at 41 of 63 farms (65%). This was shortly followed by Ross (1988) who found predation at 92 of 96 farms (96%). Work in this area has since been intermittent in Scotland. The Fisheries Research Service (2001) found that, of the 195 sites that responded to a questionnaire, 81% reported predation from seals (both grey and harbour), and Northridge et al. (2010) reported seal predation at 61 of 83 sites (73%). The problem has also been recognised in other parts of the world, probably existing wherever there is overlap between the aquaculture industry and pinniped populations. Literature discussing pinniped depredation, and at fish farms more generally, relates to industries in Australia/Tasmania ( e.g. Pemberton and Shaughnessy, 1993), British Columbia ( e.g. Olesiuk et al., 2010), Maine (Anon., 1996), Bay of Fundy (Jacobs and Terhune, 2002), New Zealand (Kemper et al., 2003), Turkish Aegean (Guclusoy and Savas, 2003) and in Chile (Sepulveda and Oliva, 2005). This range of studies shows that the problem is of global concern to the aquaculture industry, and reveals that a certain amount of disparate research effort has been invested in attempts to find and develop potential solutions.

The total financial impact of seal depredation is not straight-forward to estimate and consequently few reliable figures are available. Even a direct count of reported mortality may not provide a true estimate of the real economic impact. Some of the depredated fish may have been sickly and less likely to survive due to a predatory preference for 'easy targets' while, conversely, the effects of seal depredation may extend beyond direct loss of fish by impacting on fish growth and feeding rate (see below).

Hawkins (1985) found that while most sites experienced relatively low level predation (<£1000 per annum at 20 of 41 sites), 5 sites reported losses of more than £10,000, and one site estimated loss at over £100,000. Ross (1988) found slightly higher figures; the annual loss across 45 sites where interviews were conducted averaged £31,000 per site, with the highest reported loss per site at £280,000 per annum. Northridge et al. (2013) examined industry data from 87 sites over a 129 month period and found that almost 1.4 million fish were reported lost to seals. If these fish had made it to harvest size of 5kg with a market value of £3.5 per kg; they would have been worth almost £25 million, a loss of roughly £26,000 per site per annum. The highest reported monthly predation loss was 70,000 fish - assumed to be associated with a major containment breach or other catastrophic incident.

KG No.	Knowledge Gap
1	The extent and monetary cost of seal depredation at Scottish fish farms is unknown.

In New Zealand, one salmon farming company estimated losses of NZ $3500 - 5700 per day (~£1500 - 2500) during August 1997 (Stewardson and Cawthorn, 2004). Rueggeberg and Booth (1989) estimated that predation by seals and sea lions in British Columbia accounted for the loss of around 1% of total production. In the Pacific Northwest the problem was estimated to be as much as 10% of production cost (Moore and Wieting, 1999). The annual value of depredated fish in the Tasmanian Atlantic salmon aquaculture industry was estimated at AUS $11.5 million in 2000 (Anon., 2002a). Losses of salmon from farm sites in Maine, USA, attributed to seals between 2001 and 2003, ranged from 0 to 27,629 individual fish per site (Nelson et al., 2006), and in Los Lagos, Chile, Sepulveda and Oliva (2005) found that 90% of salmon farms had reported attacks by South American sea lions ( Otaria flavescens). Clearly then, the problem can be considered as financially significant to the industry worldwide.

2.1.2 Types of Damage

2.1.2.1 Equipment Damage

In general, damage by seals to fish farm equipment involves holes being made in the main, or growth, net. The base of the net is usually considered most vulnerable to seal attack. However,{Northridge, 2012 #456@@author-year;Thistle Environmental Partnership, 2010 #467} Thistle Environmental Partnership (2010) found that holes are found in all parts of nets, including side panels, and are more common on the base and in areas of increased wear such as near the water-line.

Damage to netting attributed to predators was the second largest cause of escapes from Scottish farms over the period from 2009 to 2012 (Northridge et al., 2013). While this form of damage is by no means insignificant - facilitating the release of at least 21,000 farmed salmon in 2011 (Northridge et al., 2013) - it is clear from previous reports that a far greater number of fish are bitten through the meshes of nets without causing escapes. In other parts of the world, particularly where Otariid species are the main cause of depredation, damage to nets may be a more serious concern. For example, Pemberton and Shaughnessy (1993) observed 150 holes which had been created in one net over a single night at a Tasmanian farm. These authors compared the different materials used to construct nets and found that two types ( e.g. 5mm diameter braided polyethylene and steel mesh), were not damaged during the course of the study. We understand that certain new types of netting material have been deployed and tested in Scotland, including high modulus polyethylene ( HMPE) - such as Dyneema ^TM , PVC coatings (such as Aquagrid ^TM ) and steel or copper cores, and that many of these new materials are marketed as having 'predator resistant' properties. While some of the trials have been reported on manufacturers and fish farm company websites, no detailed or independent assessment of their efficacy has been published.

KG No.	Knowledge Gap
2	What effect do different netting materials have upon seal depredation of salmon?

2.1.2.2 Physical Damage to Fish

The most obvious form of damage caused by seals is the direct injury of fish which can take several different forms. Northridge et al. (2013) defined four separate categories of damage inflicted by the predatory behaviour of seals at fish farms. These could be summarised as:

1) Heads - only the head of the salmon is left, where the seal has chewed and eaten the rest of the carcass.

2) Halves - half of the fish has been bitten off, leaving the anterior half intact. At one site at least these fish appeared to have been 'sucked', tail-first, through the net.

3) Gashes - multiple rake-marks or gashes with irregular spacing which does not seem to be consistent with being caused by a seal's teeth.

4) Abdominal bite - the most readily recognisable form of damage, typically found on larger fish (>1.5kg), a pair of parallel teeth marks on either side of the abdomen, just behind the gills.

It is likely that other categories of damage could be distinguished.

By categorising damage types in this way it is possible to compare and contrast their characteristics, and learn more about the mechanism of attack. For example, category 4 damage appeared to always be inflicted from below the fish, suggesting that these attacks occur through the netting at the base of the net. Category 2 damage however, was, in one instance at least, reported to have occurred through the side netting, with fish being observed by divers still stuck through the mesh. Category 3 damage seems to be caused by the seal's flippers rather than its teeth and could therefore represent a very different predatory strategy.

Different types of damage and different sizes of fish may have different likelihood of being classified and reported as seal or predator damage. Small fish for example tend to decay more quickly than large fish, which could lead to an artificially reduced rate of damage reported among fish early in the cycle. The Code of Good Practice for Scottish Finfish Aquaculture (2010 section 5.3.5.5) states that likely cause of death should be determined by a competent person, and we understand that certain companies provide training in order to meet this criteria, but there is no industry-wide standard for classification of fish mortality (known as 'morts').

KG No.	Knowledge Gap
3	Exactly what has been - and what should be - classified as seal predation mortality?

2.1.2.3 Growth Reduction

In addition to the direct loss of fish due to predation mortality, reduced growth rates due to the presence and aggressive behaviour of seals are believed to further reduce profitability (Schotte and Pemberton, 2002). This relationship is difficult to quantify and although this is a widely held perception within the industry, we are not aware of any studies which have investigated the effect directly. Schotte and Pemberton (2002) point out, however, that Atlantic salmon are thought to 'habituate' to the presence of divers, and the same process may occur if seals could be effectively prevented from causing direct damage to fish ( e.g. by anti-predator netting) or at sites where seals are present but do not regularly attack nets. During a trial of the efficacy of ADDs by one of the manufacturers at a Scottish fish-farm site, the 'Specific Feed Rate' was recorded and used as a proxy for growth rate (Ace-Hopkins, 2002a). This parameter was compared with the number of 'seal detections' measured by the trigger devices of the Ace-Aquatec Silent Scrammer acoustic deterrent device. The industry report claims that predation by seals significantly reduced the 'Specific Feed Rate' and they hypothesised that this would result in reduced growth rate. One concern with this study is that triggers do not measure seal depredation directly, instead they are triggered when they are agitated by the movement of the fish. Seal depredation can cause fish to panic and knock the triggers, however, farm managers have reported that the triggers can also be activated by other forms of fish movement and indeed by poor weather. It is possible that both of these causes could be correlated with changes in feeding rate. Modern fish farms control and regulate food delivery by monitoring feeding rate. It should therefore be quite straightforward to look for a correlation between food delivery, feeding rate and direct measures of seal presence at cages and depredation using data the industry has already collected.

KG No.	Knowledge Gap
4	How are salmon growth rates affected by seal presence and depredation?

There are also anecdotal reports from the Scottish industry that increased stress levels caused by depredation, or attempted depredation, make fish more susceptible to disease (Northridge et al., 2010). Nash et al. (2000) refer to outbreaks of Hitra, a bacterial disease, starting in and having greatest impact on pens which were already being attacked by seals. It is known that individual seals sometimes travel between fish-farm sites in Scotland (Northridge et al., 2013), and there is therefore a possibility that seals may act as a vector for disease (Ross, 1988). This possibility has not to our knowledge been investigated and the suggestion is speculative only.

KG No.	Knowledge Gap
5	Is there a relationship between seal depredation and disease among farmed salmon?

2.1.2.4 Stress/Welfare Concerns

In addition to the possibility of illness from elevated levels of stress, there are more direct welfare issues in cases where farmed fish are injured but not killed by seals. Farmers clearly have a duty of care to protect their stock and minimise suffering in this situation. The recent licensing system for predator removal in Scotland regulates applicants who have a need to protect the health and welfare of stock. There were 39 applications for licences to shoot seals on the basis of fish welfare concerns in 2011, of which 32 were granted.

Again, to our knowledge this type of damage is unquantified, but the number of fish injured in this way could feasibly be assessed by farm managers or by observations from fish farm video monitors.

KG No.	Knowledge Gap
6	Quantification of welfare concerns - to what extent do seals injure without killing fish?

2.1.3 Mechanism of Attacks

In order to manage seal depredation and design techniques and protocols to reduce it, it is important to understand seal behaviour during depredation events. There is remarkably little literature documenting the methods of seal predation on farmed fish. There is however anecdotal evidence available from fish farmers, some of which have been reported previously. Here we describe some accounts most relevant to the subject of this report.

Tillapaugh (1991) [as described in Ace-Hopkins (2002b)] reported an attack by a harbour seal ( Phoca vitulina) in Canada as witnessed by a diver, " The seal circled the netpen until the fish were frightened enough to charge the opposite side of the pen. The seal then dove under the pen and attacked the fish pushing against the side of the netpen. This procedure was repeated several times." Ross (1988) also described seals 'charging' at the net, causing fish to panic and the seal was then able to grasp the fish inside a fold of net.

Ace-Hopkins (2002b) described harbour seal attacks, as witnessed by farm workers, in greater detail, " Common seals rarely damage the growing net but grab a fish between their front paws, bite the abdomen of the fish and suck the guts through the mesh" and, " When nets are loosely tensioned farm workers have also reported that common seals can manipulate the growing net into a pocket directly, so entrapping a fish that swam too close". There is further evidence to suggest seals may take advantage of pockets of slack by manipulating netting. Iwama et al. (1997) described harbour seals, again in Canada, creating a pocket in the netting and thus entrapping a fish.

On the attack methods of grey seals, Ace-Hopkins (2002b) described two further methods of attack; " Apparently one grey seal was able to "climb" the 3 feet to the handrail and squeeze between the two nets and thus swim with the fish. No one witnessed his entry but did witness his exit (on three occasions). The animal was seen to swim at speed to the exit point, use his impetus to leap from the water, rotate and arch its back in the manner of a human high jumper (the Frisbee flop [sic.] ) and landing in the sea in one fluid movement." And, " In strong tides the growing nets become distorted and the fish inevitably swim closer to the nets than they would otherwise do. The grey seal hooks itself onto the upstream side of the net and waits for a salmon to come too close. When the seal judges the fish is within range he increases his drag by letting go with his back flippers (becoming more upright) and using this impetus coupled with his strength to make the growing net into a pocket to entrap the fish."

Where predator nets (a secondary net positioned outside of the main net) are used and the gap between the two nets is insufficient, Iwama et al. (1997) stated that seals can manipulate both nets simultaneously. A video available online ^[10] shows a sea lion at an unknown location pushing an outer anti-predator net in order to reach dead fish in the bottom of the main net.

Iwama et al. (1997) also suggested that harbour seals looked for small openings in anti-predator net systems where these were being used. This mechanism has been observed by previous work in Scotland (Northridge et al., 2013) where footage was captured of grey seals in between the growth net and the anti-predator net. In this way the seal only needs to manipulate one layer of netting in order to grasp the fish, but clearly there is an increased risk of the seal becoming entrapped between the two layers of netting.

Anecdotal evidence from Schotte and Pemberton (2002) described several methods of attack employed by fur seals, including: corkscrewing through anti-predator netting side panels, pushing the base of anti-predator netting upward into the growth net and gaining access between growth and anti-predator netting at the surface. Predation here was thought to be more pronounced at times of high tidal flow, and fur seals were thought to seek the easiest opportunity to access fish (which may be the next pen, or the next farm site). These behaviours seem to broadly correspond with those reported from phocid seals elsewhere.

In some locations at least, pinniped depredation is known to involve groups of animals. Pemberton and Shaughnessy (1993), for example, recorded 12 attacks (of 106 in total) where more than five fur seals were recorded. Tillapaugh et al. (1993) also observed predators working in groups. In other locations, however, this appears to be uncommon; Northridge et al. (2010) found that single 'rogue' animals were reported to be the cause of the majority of predation (61 of 83 sites responding to a questionnaire), and Guclusoy and Savas (2003) reported that in 38 of 40 monk seal ( Monachus monachus) attacks recorded at fish farms in the Turkish Aegean only a single animal was involved.

KG No.	Knowledge Gap
7	What specific mechanisms do pinnipeds use to damage fish within nets?

2.1.4 Potential Solutions Overview

Several research groups have considered potential solutions to the issue of pinniped depredation on marine aquaculture sites worldwide. In the main, their publications have taken the form of review articles, often written for the audience of their respective regulatory body. Specific points of interest from these reviews will be presented in more detail at relevant points in this report; here we provide a brief overview of each of these reviews to illustrate the extent of previous work in this area. We have included research relating to species not found in Scotland because to exclude this work would be to reduce the available information significantly, and in many cases the findings are directly comparable with the Scottish situation.

Ross (1988) discussed various methods for predation control in Scotland. The main mitigation techniques considered were the use of anti-predator nets, ADDs and shooting, reflecting industry practice at the time. The negative impact of anti-predator nets, as perceived by the industry, was found to vary significantly between locations. Fouling, reduction of water flow and entanglement of predators were the main problems cited. Entanglement and drowning of animals was considered unacceptable at some sites, but at others it was considered justifiable or even desirable.

Arnold (1992) reported to Greenpeace on attempts to improve the predator exclusion measures to protect aquaculture sites in Shetland. The author discussed the use of various deterrence and exclusion methods including anti-predator netting, lethal removal and acoustic deterrents, but the emphasis of the report is on improved techniques for weighting and net tensioning systems, particularly the benefits of the use of 'sinker tubes' which were relatively new to the industry at the time. The continued development of these to improve predator exclusion, along with other net tensioning methods, was considered to merit serious investigation.

Pemberton and Shaughnessy (1993) reviewed and assessed the methods of deterrence used for Australian fur seals ( Arctocephalus pusillus doriferus) in Tasmania, including ADDs, lethal removal, pursuit with boats, floodlights activated to scare seals and emetics. The authors concluded that physical exclusion was the 'best' solution and that shooting was ineffective.

Smith (1994) provided a useful literature review for the Canadian Department of Fisheries and Oceans, summarising much of the very early work, including that on acoustic deterrence. He noted agreement among several authors that current acoustic deterrents were expensive and often ineffective. He recommended an emphasis on the prevention of predation and the use of properly designed and maintained anti-predator nets was seen as the most effective way of doing this. The use of emetics (discussed further in section 2.3.3) was described as worthy of further research having shown some effectiveness in limited trials. Harassment techniques, such as pursuit of problem animals, and the use of lights, explosives and warning shots were not found to have lasting effectiveness.

The report of the Gulf of Maine Aquaculture-Pinniped Interactions Taskforce ( NMFS, 1996) described efforts made to mitigate perceived problems with pinniped conflicts in the Maine aquaculture industry. The taskforce was partly established to address recent changes to the US Marine Mammal Protection Act ( MMPA) which came into force in 1995 and prohibited the use of lethal force to control predators. The use of explosives (seal bombs and cracker shells) was considered to have some benefit to the industry, when used responsibly. Concerns over the use of acoustic deterrents were the expense ( US $10000 - 12000 per system) and the likely effects on non-target species. They were, however, considered to be a useful tool for growers, and authors recommend that they should be available for use. Deployment of predator models and playback of vocalisations (killer whales) were described as having very short-term effects only, but it was suggested that these could be useful in some scenarios. The use of live marine predators (killer whales and sharks) was dismissed as impractical. Aversive conditioning, including shock-collars and emetics administered through dart-injection or food bait, was deemed to be worthy of further study. Translocation of problem individuals was also discussed, but was thought to be prohibited under deterrence regulations, and was not considered further. The use of boats to harass nearby animals was described as potentially useful, as was the presence of humans on the pen-site. The use of dogs on sites was mentioned but not considered to be useful. Anti-predator nets in their different forms ( e.g. 'curtain' - one sheet of netting surrounding a net without a base, 'box' - a box of netting which encloses a growth net and 'perimeter' - a large net surrounding the entire site) were found to be of some use, but limited by the technical challenge of achieving adequate weighting and maintaining a useful distance between the predator and growth net. Good husbandry practice in the removal and disposal of fish mortalities was also highlighted as a potentially important factor.

A workshop was held in Seattle, USA, in 1996 to consider the problems and uncertainties surrounding the use of acoustic deterrents in commercial and conservation practices. The proceedings (Reeves et al., 1996) included a discussion of the use of ADDs to mitigate salmonid predation by pinnipeds. Consistent aversive effects were only reported at very high sound intensities. It was suggested that resistant individuals might have had impaired hearing and/or have learned avoidance behaviour or habituation.

Iwama et al. (1997) reviewed some previous research surrounding the use of ADDs at commercial net-cage salmon farming in British Columbia, Canada, as well as alternative methods of deterrence/exclusion. The authors recommended the prohibition of ADDs because there was so little evidence that they were effective and made several recommendations regarding physical changes to cage systems such as pen shape, mesh size and net flexibility.

Moore and Wieting (1999) reported on a US National Oceanographic and Atmospheric Administration ( NOAA) workshop addressing interactions of aquaculture with marine mammals and turtles, including a discussion of acoustic deterrence and the likelihood of habituation. The principle concern of the report was the improvement of industrial practice in response to rapid growth of marine mammal populations. The main areas highlighted were: engineering improvements to cage-design and anti-predator nets, development of more effective acoustic deterrents, relocation of sites offshore, relocation or elimination of 'rogue' animals and reduction of local populations through reintroduction of pinniped harvest programs. They made many recommendations for further research including the need for characterisation of marine mammal interactions and behaviour around aquaculture sites and the investigations of new net technologies.

Nash et al. (2000) discussed the extent of pinniped depredation at aquaculture sites in the Pacific NW and potential solutions, including relocation of problem animals and the use of ADDs, both of which they dismissed as being valuable as short-term strategies only. The only long-term solution suggested by this report was the relocation of fish-farm complexes away from haulout sites.

An analysis of the predator control techniques used in British Columbia by Jamieson and Olesiuk (2001) described all harassment techniques (explosives, acoustic deterrents, 'tactile harassment' and chasing by vessels) to be ineffective in the long-term. Anti-predator netting was described, but no details of efficacy were given. The risk of entanglement and drowning, as described elsewhere, was noted. Bio-fouling of nets was thought to reduce the incidence of predation, possibly due to the net having reduced pliability or by reducing the predator's view of the fish. Translocation of problem animals was thought to be ineffective because the animals often returned to the capture site. The authors noted that the existing government guidelines discouraging the location of sites closer than 1km to a seal haulout had no scientific basis, and the authors note that harbour seal telemetry data indicated daily foraging movements in the range of 10km.

Würsig and Gailey (2002) reviewed various aspects of marine mammal interactions with shell-fish and fin-fish aquaculture worldwide. They categorise deterrent techniques into six major categories; (i) harassment; (ii) aversive conditioning; (iii) exclusion; (iv) non-lethal removal; (v) lethal removal; and (vi) population control, and provided a short review of each. They concluded that predator interactions need to be considered from the start of an aquaculture site installation so that effective solutions can be factored into the cost of the facility, rather than hoping for quick fixes later on. The methods most likely to provide functional long-term solutions were; exclusion of predators through physical barriers, non-lethal removal of problem individuals and aquaculture facilities being located further from known haulouts.

A 2002 report by the Tasmanian Marine and Marine Industries Council (Anon., 2002a), summarised the anti-predator techniques from aquaculture industries worldwide and provided a comprehensive review of methods employed in Tasmania. Mitigation methods considered were acoustic deterrents (including explosives), capture and relocation of problem individuals, improved exclusion techniques such as anti-predator netting, tactile harassment (rubber bullets and cattle prods), chasing of animals by vessels, taste aversion, electric fencing (to prevent seals climbing across walkways), lethal removal, population control (culling) and the use of a device which emits an electric field to repel sharks (not considered worthy of further investigation). Their discussion of tensioning methods for anti-predator netting suggests that this is an important mitigation technique for Tasmanian farms. Specific methods for tensioning Australian aquaculture nets are described in detail in another review by Schotte and Pemberton (2002) (see section 2.3.1 for more details). Acoustic deterrent devices were characterised as having 'limited effect'. Airmar devices have been trialled at Australian tuna farms and were reported to have 'mixed success', with farmers believing that any apparent effect disappeared after a year, after which a 'dinner bell' effect was reported. Capture and relocation was not thought to be an effective long-term strategy due to the cost, risk of disease transmission, and ethical issues associated. The authors of both reviews concluded that no easy or fool-proof method for mitigating interactions was available, stressing the need to more effectively manage inevitable interactions, rather than trying to 'solve' the problem.

Petras (2003) reviewed potential deterrence measures for reducing killer whale predation on Steller sea lions ( Eumetopias jubatus) near the Western Aleutian Islands. The author focused on acoustic deterrents, and discussed both pingers and seal scarers in detail. The use of existing devices in this context was concluded to be speculative at best, and the need for behavioural research was stressed.

Guclusoy and Savas (2003) discussed and compared techniques used for deterring Mediterranean monk seals from predating on gilthead sea bream ( Sparus auratus) and European sea bass ( Dicentrarchus labrax) at fish farms in the Turkish Aegean. Farmers tried flashing lights at seals, feeding them with pesticide-injected fish, underwater noise (banging the walkway or tin cans) and both warning and direct gunshots, all of which were reported to be unsuccessful. Anti-predator netting was used at 6 of 25 sites, but all reported difficulties in creating an effective barrier. Seals found gaps in between curtains of netting, or in the case where the net extended to the seabed, they found gaps where insufficient sinkers had been installed on the ground rope. Authors later supervised the adjustment of anti-predator netting, after which no more losses were reported (unfortunately there are no details reported as to the changes effected).

Baird (2004) and Stewardson and Cawthorn (2004) reviewed the use of deterrents including ADDs, Pulsed Power Devices ( PPD - which generates an underwater shockwave), predator noises, gunshots, pyrotechnics, taste/scent deterrents, tactile deterrents and vessel chasing against fur seals in New Zealand aquaculture and fisheries. None of the acoustic deterrents reviewed were found to have sustained effectiveness, but further research was recommended into the potential of ADDs and taste deterrents (see section 2.3.3).

Nelson et al. (2006) used a modelling approach to analyse the influence of farm siting and ADD use on seal depredation rates at fish farms in Maine, USA, between 2001 and 2003. Siting apparently had a significant effect on depredation rate, with farms further from haulouts being less affected. They found no evidence that ADD use reduced seal depredation.

Robinson et al. (2008a) and Robinson et al. (2008b) reviewed the practice of fur seal relocation from around Tasmanian fish farms in detail. The methodology appeared to be well developed and frequently used, but despite this the authors concluded that it only provided short-term relief from depredation (see section 2.3.5).

The results of questionnaire surveys of fish farm managers on seal depredation and management at fish farms in Scotland and their apparent relative efficacy, are provided by Quick et al. (2002), Quick et al. (2004) and, more recently, by Northridge et al. (2010) and Northridge et al. (2013). Generally, these reports showed that net tensioning was believed to be the key factor in minimising depredation events. Northridge et al. (2013) also examined industry data, and found that farm sites located in closer proximity to seal haul out sites did not experience higher seal damage levels.

A report from 'Hydroacoustics Incorporated' (De La Croix, 2010) compared several different varieties of acoustic deterrent devices, including explosives, ADDs and 'pulsed power' devices. It also provides some consideration of their relative merits, with explosives and ADDs being found to have limited short-term effects only. An impulsive airgun device which the company ( HAI inc.) was marketing for use at fish farms was described but this was yet to be tested in a real-world scenario.

The extent of conflicts between aquaculture and marine mammals in the Southern hemisphere is reviewed in Kemper et al. (2003), particularly addressing finfish aquaculture in South America, Australia and New Zealand. This article addressed the methods of deterrence used at various aquaculture operations and their varying degrees of success. They summarised other reports on anti-predator methods in the Southern hemisphere, and found no empirical evidence for the efficacy of ADDs. Anti-predator netting was reported to be effective at some locations; however, lethal entanglements were also reported. The characteristics of anti-predator nets which lead to entanglements were: too large a mesh size, unrepaired holes, nets not enclosed at the bottom, loose and baggy nets and inappropriate feeding practices which encouraged marine mammal interactions. At Marlborough Sounds, New Zealand, where anti-predator nets are enclosed at the base and made from stiffened nylon, there had been no recorded entanglements (there was no further detail given of the study).

A review paper by Scordino (2010) reviewed efforts made by the National Marine Fisheries Service ( NMFS) West Coast Pinniped Program to reduce salmonid predation by harbour seals and California sea lions in rivers and estuaries. This included a detailed assessment of the large number of techniques trialled: above water and underwater explosives, pulsed power devices, taste aversion, predator models, chasing by vessels, rubber bullets, physical barriers, electric barriers, capture and relocation, population control, lethal removal of problem individuals and acoustic devices including predator noises. The general conclusion was that non-lethal measures have had limited effectiveness. Work at the Ballard Locks, Seattle, over many years had shown that in order to consistently cause aversion it was necessary to inflict physical pain. Otherwise, the only effective solutions had involved the removal of problem animals.

A technical review of the noise associated with marine aquaculture in Canada (Olesiuk et al., 2010), included a discussion of acoustic deterrents and explosives/pyrotechnics used to mitigate pinniped depredation. The focus of the report was the likelihood of detrimental effects on target and non-target species. Acoustic deterrent were described as only being effective at deterring naïve seals and the authors suggested that benefits were minimal.

Pinniped interactions with aquaculture, and techniques for managing them, have been addressed by a number of authors in many different locations and contexts as summarised above. None have found evidence that any one method can provide an effective solution. Many suggest that a suite of anti-predator methods will be necessary in most situations, and several emphasise the need for anticipating the likelihood of predator interactions from the early planning stage so that mitigation can be factored into the cost of the facility from the start. The need for improved exclusion techniques, such as anti-predator nets, is one area where further investigation is warranted. These nets have been reported to cause entanglement and drowning of birds and seals at some sites (including many in Scotland), however several authors report that acceptable solutions to these problems were found in some locations (Anon., 2002a; Jamieson and Olesiuk, 2001; Kemper et al., 2003). No authors have shown convincing evidence for the long-term efficacy of acoustic deterrents. Many suggest the need for further research into the effects of ADDs on both target and non-target species (see also section 2.2.3). Emetic and conditioned taste aversion techniques have shown promise and have been described as being in need of further research in several reviews (see section 2.3.3).

2.2 Acoustic Deterrent Devices to Prevent Depredation

2.2.1 Types of ADDs in Use and Characteristics

Table 2 summarises the acoustic characteristics of the devices most frequently used in Scottish fish farms, but it should be noted that a variety of devices has existed, many of which have had ephemeral usage. Of these, interview surveys suggest that Airmar, Terecos and Ace Aquatec are most widely used in Scottish aquaculture (Northridge et al., 2010). Where possible we have provided both the manufacturer's figures, and independently obtained field measurements that in some cases differ substantially from those stated by the manufacturers, indicating considerable uncertainty about the actual source levels of the devices. All measurements in the following are dB re 1 µPa @ 1m. Amplitude measurements are usually taken as either:

• 'peak to peak' (the amplitude difference between the most positive and the most negative excursions of a signal, over a given time period);
• 'zero to peak', or 'peak' (the amplitude of the greatest excursion from zero over a given time period);
• or Root Mean Squared, or ' RMS' (the square root of the mean of the square of the signal from zero over a given time period).

Unfortunately, this key piece of information for comparing source level measurements is often overlooked, and we have marked these instances below with 'Unknown'.

One of the most commonly used devices is the Airmar dB Plus II and a range of sources levels have been reported for this. The manufacturer's manual provides a source level of 198 dB ( RMS) but field measurements have differed widely.Jacobs and Terhune (2002) measured Airmar ADDs in the Bay of Fundy, Canada, and found source levels of only 178-179 dB (peak to peak), while Haller and Lemon (1994) reported higher values at 183 dB ( RMS) (and 194 dB RMS when looking at individual pulses). Lepper et al. (2004) reported a source level of 192 dB ( RMS), while most recently Brandt et al. (2012b) estimated the Airmar source level as 190 dB ( RMS), with peak pressure level of 206 dB.

The manufacturers of the Lofitech device provide a source level of 189 dB (unknown), but most field measurements have suggested a higher source level. Yurk and Trites (2000) reported maximum SPL as 194 dB (unknown) and Shapiro et al. (2009) 193 dB ( RMS). Brandt et al. (2012b) calculated a source level as 194 dB ( RMS) with a peak pressure level of 205 dB and Westerberg et al. (1999) measured 191 dB (peak to peak) source level. Measurements by Graham et al. (2009) matched the manufacturers' specification of 189 dB (unknown). By contrast, Fjalling et al>. (2006) measured a Lofitech device as having source level of just 179 dB ( RMS).

The Terecos is one of the least powerful devices used routinely at Scottish aquaculture sites. The manufacturers do not provide a reliable source level, however Olesiuk et al. (2010) report the source level of a Terecos DSMS-4 to be 185 dB (unknown), whereas Lepper et al. (2004) found the same device to have maximum SPL of 179 dB ( RMS).

The Ace-Aquatec Universal Scrammer has a source level of 194 dB (unknown) according to the manufacturer, which corresponds well with measurements of 193 dB ( RMS) made by Lepper et al. (2004).

The large discrepancies in source levels are notable. Some of the lowest values, such as the Airmar source levels reported by Jacobs and Terhune (2002) may result from faulty or incorrectly configured equipment. Other discrepencies probably reflect uncertainties in the way in which measurements are made. For example, for pulsed and intermittent sounds, RMS levels depend critically on the time window over which mean values are calculated. In addition, the frequency ranges over which measurements are made are rarely reported. From the perspective of assessing the possible effects of these devices on auditory systems, sound exposure levels ( SELs) and peak pressure levels will usually be the more relevant acoustic measurements, yet these values are rarely if ever presented.

Generally, manufacturers have not provided (nor been required to provide) data that adequately describe the acoustic output of their devices in a manner that would allow an assessment of effects on both target and non-target species to be made. Many organisations, including the OSPAR commission, now recognise that underwater noise is a form of pollution (Gotz et al., 2009). From this perspective, the dichotomy between the required levels of monitoring regarding chemical and acoustic pollution is striking.

KG No.	Knowledge Gap
8	Exact acoustic output of all devices and an appropriate metric (or suite of metrics) for comparison of different signal types.

To our knowledge, there is a maximum of five devices which are currently employed in Scottish aquaculture. These are summarised in Table 2. One of these, the Ferranti-Thomson, is no longer produced, but we believe it may still be used by a small number of sites.

Table 2 Acoustic Characteristics of Acoustic Deterrent Devices Used at Scottish Aquaculture Sites

Manufacturer	Device	Source Level ( dB)		Frequency (kHz)	Reference
		Scientific Literature	According to Manufacturer
Airmar	dB Plus II	192 ( RMS)	198 ( RMS)	10 (tonal - with harmonics)	Lepper et al. (2004)
Lofitech	Universal Scarer	193 ( RMS)	189 (Unknown)	14 (tonal - with harmonics)	Shapiro et al. (2009)
Ace Aquatec	Universal Scrammer 3	193 ( RMS)	194 (Unknown)	10 - 65 (broadband)	Lepper et al. (2004)
Terecos	DSMS-4	179 ( RMS)	None given	2 - 70 (broadband)	Lepper et al. (2004)
Ferranti-Thomson	4X	166 (Unknown)	200 (Unknown)	7 - 95 (broadband)	Terhune et al. (2002)

Figures 1 to 4 show the spectral characteristics of Ace-Aquatec, Terecos, Lofitech and Airmar devices (our own unpublished work; Gordon and Northridge, 2002). All devices have high frequency (ultrasonic) components to the sound signal, but only the Lofitech device could be seen to exceed ambient noise levels above 100 kHz. One particular harmonic band from the Lofitech sits at c. 120 kHz, in the same frequency band as the echolocation clicks of the harbour porpoise, raising the potential for masking of echolocation/communication behaviour.

It is worth noting that the of sensitivity of grey and harbour seal hearing reduces dramatically above ca. 40 kHz (see audiograms in Gordon and Northridge, 2002), and therefore higher frequency noise created by acoustic deterrents is effectively unnecessary.

One suggested explanation for temporary lack of efficacy from seal scarers has been low source level due to the build-up of marine fouling on the transducer elements (Olesiuk et al., 2002). However Northridge et al. (2013) found no increase in source level after cleaning very severe fouling from the transducer of a Terecos ADD at a Scottish salmon farm. The dominant fouling organisms in this case were sea squirts (which are largely water), with a relatively juvenile community of calcified organisms such as mussels, scallops and barnacles. Further colonisation of the transducer by these or other hard shelled fouling organisms may have a larger effect on source level.

Voltage drop has also been cited ( e.g. Gordon and Northridge, 2002; Olesiuk et al., 2002) as one possible cause for occasional inefficacy of devices, but to our knowledge no study has yet demonstrated the output of the three most common devices under reduced voltage. Harris (2011), working with a Lofitech device, reports finding evidence of a 1.5 dB (presumably re 1 µPa, RMS or Peak) decrease in output signal correlating with a voltage drop of 2.6 V (from 12.5 to 9.9 V). This relatively low drop in sound output indicates that this model at least is quite robust to voltage drop. Clearly this relationship between voltage and output level could usefully be explored in other models too.

KG No.	Knowledge Gap
9	Effect of fouling and voltage drop on signal output (under full range of operating conditions).

Figure 1 Waveform and Spectrogram of an Ace-Aquatec US3 (70 kHz LP filter)

Figure 2 Waveform and Spectrogram of a Terecos ADD - program 4 (70 kHz LP filter)

Figure 3 Spectrogram of a Lofitech ADD, showing harmonics up to c. 150 kHz

Figure 4 Waveform and Spectrogram of an Airmar dB Plus II (1 kHz HP filter)

N.B. Recordings made with PAMGUARD software, using B&K 8103 hydrophone (flat response up to 125 kHz). Analysis through Raven Pro 1.4 (Cornell Lab of Ornithology) with 2048 point FFT and 2000 sample Hamming window. Axes vary between figures.

2.2.1.1 Duty Cycles

The duty cycles of the three main devices employed in Scotland are documented by Lepper et al. (2004), who looked in detail at the acoustic properties of the Airmar, Ace-Aquatec and Terecos devices. They stated that:

"The Airmar system has a 1.4 ms tonal burst with 40 ms spacing. The sequence is repeated with a 50% duty cycle allowing an approximate 2s quiet period." This device has the ability to operate in a 'low-power mode', where the duty cycle is reduced from 2.5s ON - 2s OFF, to 2.5s ON - 6.5s OFF (Airmar Owner's Manual). The manufacturer states that the device should not be left in this mode for long periods, as it will ' result in less than optimal protection from predators'.

"The Ace-Aquatec has a randomised sequence with a 50% duty cycle for a 5 s period. The relative length of pulses uniformly shortens from 14 ms to 3.3 ms followed by a shift in frequency of the tonal components and their equivalent distribution to each other."

"The Terecos has four different programs. Program 1 is a sequence of repetitive five segment (16 ms duration) continuous tonal blocks forming an up and down frequency sweep. Program 2 was a randomly timed sequence of continuous and time variant multi-component tonal blocks. Program 3 consists of sequences (Seq.2) of eight segment (8 ms duration) continuous tonal blocks forming an up and down frequency sweep combined with variable continuous multi-component tonal blocks. Program 4 has a randomly timed combined sequence of Seq.1, Seq.2 tonal blocks, continuous multi-component tonal blocks and time variant multi-component tonal blocks."

We believe that Terecos manufacturers periodically make changes to programs in order to reduce the likelihood of habituation, so the programs described here may not be typical.

2.2.1.2 Triggers

Triggers would allow activation of ADDs only when a predator is detected, or when depredation is occurring. It has long been recognised that a reliable and automated triggering mechanism would probably increase the efficacy of devices and furthermore would have the potential to greatly reduce the amount of acoustic energy released into the environment (Mate and Harvey, 1986). Several reports have also called for the development of reliable triggers, including those of Gordon and Northridge (2002),, Anon. (2002a); Kastelein et al. (2000) and Smith (1994).

A triggered device was reported to have been trialled in Vancouver, Canada, as early as 1988. This detector activated an acoustic device when the nets received an erratic, sharp impact (Smith, 1994), but development of this device does not seem to have gone very far. According to Olesiuk et al. (2010), Airmar explored the development of triggers activated by sonar or detection of predator vocalizations, but these were not successful. The product sheet for the Airmar dB Plus II model states that there is an input socket designed for the attachment of a 'mammal detector', or could alternatively be operated manually by a remote operator observing an attack via net monitoring video.

The only concerted attempt that we are aware of to develop effective and useful triggers has been by Ace-Aquatec, whose triggers are designed to be activated by the movement of fish in response to a seal attack (Ace-Hopkins, 2001). When a seal approaches the net, it is expected that the fish will become agitated and this movement is detected when they collide with sensors placed inside the net. Unfortunately we are not aware of any independent studies which have looked at the efficacy of these triggers, and our own discussions with site managers have suggested that false detections are not uncommon. Northridge et al. (2010) reported that, among their interview sample, predator triggers had been tried at 27 sites in Scotland, but none of the interview sample had judged them to be successful.

2.2.1.3 Modes of Operation

There are two broad strategies for the use of ADDs at fish farms. The first is to have the device emitting sound continuously (except during diving operations), forming a continuous sound field. Northridge et al. (2010) found that at 28 of 52 sites where ADDs had been deployed, the devices when deployed were left on continuously. The rationale here is that if the predator is excluded from the immediate vicinity of the farm from the outset it will never learn to associate it with the presence of a food source and should therefore have no incentive to predate on the fish there. However, continuous operation could be more likely to lead to habituation, and negative effects on non-target species due to acoustic disturbance and/or exclusion are also likely to be greater. The alternative strategy is 'responsive', with the device only being switched on for a limited period of time in response to either seal presence or attacks. Northridge et al. (2010) found that this strategy was used at 25 of 52 sites. The benefits of this approach are that the reduced duration of sound emission may limit the potential for hearing damage or habituation of local seals and reduce disturbance effects on non-target species. However, this strategy might increase the risk of seals learning that farms provide depredation opportunities before the seal has been detected and ADDs are activated. Once an association has been made in this way, the predator may be more motivated to ignore the acoustic deterrent.

In our experience, different farm managers adopt either one strategy or the other and we have therefore never had the opportunity to directly compare the two. There has, to our knowledge, been no objective assessment of the relative effectiveness of the different approaches in the context of marine aquaculture.

KG No.	Knowledge Gap
10	What is the relative efficacy of different ADD deployment 'strategies', and how can they be appropriately compared?

A recent undergraduate thesis examined the efficacy of different ADD strategies in mitigating seal interactions with anglers on the Ythan estuary, Aberdeenshire. During each sampling occasion (a day), angling boats employed one of three strategies; no ADD, continuous ADD or responsive ADD, where the device was only used in response to seal presence. It was found that anglers were more likely to have a successful trip if the acoustic deterrent was switched on for the duration (fish caught on 50% of trips, n = 18) than when it was used responsively (fish caught on 20% of trips, n = 7). This compared to a success rate of just 5% when the seal scarer was not used at all (n = 26) (Rae, 2013). While the experimental design of this part of the study was not perfect (data were collected by the anglers themselves, and the treatment regime is not documented and could therefore have been biased), there is an indication of a difference between predatory behaviours in response to different deterrent strategies. The context here was clearly very different from that at a fish farm but it does indicate that simple experimentation could provide useful data.

In addition to the two strategies described above, Quick et al. (2002) showed that some sites in Scotland use ADDs seasonally, which could be described as a sub-strategy or refinement of the approaches described above, as one or other of the broader strategies will generally be employed during the period of ADD usage. Northridge et al. (2010) found that at 12 of 52 sites, managers had only switched on devices when fish reached a certain size at which they were deemed vulnerable and then left them on. There is very wide variation in deployment tactics, and these are usually specific to the company or even individual site.

2.2.2 Extent of Use in Scottish Farms and Elsewhere

2.2.2.1 Current Usage

At present the use of ADDs at a site is permitted or restricted by local planning authorities as part of the planning consent process. Scottish Natural Heritage ( SNH) is a statutory consultee at the planning stage and can object to the planned use of ADDs - in which case the planning authority may exclude the use of ADDs in the planning consent. At present, therefore, while no specific licence is required to use ADDs as a matter of course, a licence may be deemed necessary by SNH under the Conservation (Natural Habitats, &c.) Regulations 1994 if it is thought that their use may disturb cetaceans. Scottish Natural Heritage has objected to use of ADDs in relatively few instances so most farm sites are free to deploy ADDs, but the number of sites which have requested permission to use them, and how many have been denied, is undocumented.

KG No.	Knowledge Gap
11	How many sites have been denied approval for ADD use under planning regulations, and what criteria have been used to assess applications?

Hawkins (1985) found ADDs at 4 of 41 sites (9.7%) and soon afterwards Ross (1988) found that ADDs had been in use at 8 of the 45 sites visited (18%). This proportion continued to increase through the 1990s and Quick et al. (2002) found 52% of farms reported that they use ADDs, but that usage patterns varied greatly. These results are similar to those reported by Northridge et al. (2010) who found 40 out of 81 sites interviewed were using ADDs. Northridge et al. (2010) found that of farms with ADDs in Scotland, 42% were using the Terecos model, and 35% were using Airmar while Shrimpton (2001) found that the Airmar models represent approximately half of those in use (16 of 31) in another sample.

Elsewhere, Johnston and Woodley (1998) found 22-46% of sites in Bay of Fundy were using ADDs, while in Chile, Sepulveda and Oliva (2005) found that 33% of sites used ADDs in efforts to reduce interactions with sea lions. It is clear that ADDs are not universally considered essential.

KG No.	Knowledge Gap
12	Total extent and distribution of ADD usage in Scotland is currently unknown.

2.2.2.2 Propagation, Sound Fields and Ranges of Effects

The level at which an animal at a given range will receive the sound from an ADD depends on both the source characteristics of the device and propagation loss. Propagation conditions will vary between sites, being affected by parameters such as bathymetry and bottom type. Seasonal changes in variables such as water temperature profiles will also have an effect. However, propagation loss is reasonably well understood. It can be modelled using various approaches, there is a host of empirical data from representative sites, and it is also relatively easy to check predictions by making recordings at particular locations.

Predicting aversiveness relies on many contextual and species specific factors and is therefore much more complicated than prediction of the range of audibility. Audibility will be limited either by the hearing threshold of the animal or the ambient noise level, whichever is higher. Since hearing thresholds vary between different species, the range of audibility will be species dependent.

A detailed investigation of the potential sensitivity of marine mammals to acoustic deterrents at close range was conducted by Lepper et al. (In Review). In this report, appropriate models were used to generate lookup tables of propagation loss to apply within 500m of Scottish farm sites based on characteristics such as water depth, slope and bottom type. Simple geometric models of sound propagation have been shown to be inadequate for predicting the complex sound fields (Shapiro et al., 2009) typical of relatively shallow-water environments where fish farms are generally sited. Certain key parameters were therefore tested as predictors of propagation loss including: source amplitude and spectral characteristics, water depth, sediment type, seabed slope and surface roughness. The typical frequency range of ADDs in use in Scotland is 2-40 kHz, and this range was subdivided by Lepper et al. (In Review) into frequency bands one third of an octave wide (an octave being a doubling of frequency) in order to assess frequency dependent propagation. Other than general noise propagation, this work has particular importance for estimating the risk of hearing damage to both target and non-target marine mammal species (discussed below in section 7.2)

Predictions by Lepper et al. (In Review) showed a reasonable fit to empirical data collected during earlier studies (Booth, 2010; Northridge et al., 2010). Lepper et al. (In Review) used predicted noise fields for different devices to explore the potential for hearing damage at these sites for both seals and small cetaceans. The extent of this potential risk was highly dependent on the animal's behaviour and movement within the sound field but it was evident that if animals do spend extended periods close to ADDs, SEL thresholds for permanent hearing damage based on Southall et al. (2007) would be exceeded.

To make a crude estimate of the marine area that might potentially be affected by ADD usage, we have plotted areas based on the estimated range of effects around all of the licensed fish farm sites in Scotland (figure 5 & 6). Jacobs and Terhune (2002), using a mixed model of cylindrical and spherical spreading loss, calculated the theoretical maximum range of detection to a harbour seal (higher hearing threshold than a harbour porpoise) to be 20.2 km for an Airmar device. Using median levels of ambient noise, the zone of audibility was calculated to be 9.7 km. Brandt et al. (2012b) also stated that a loud acoustic deterrent (such as an Airmar) could be audible to a harbour porpoise at a range greater than 20 km. This was based on a lower rate of transmission loss in their study area (possibly due to water depth and/or bottom type), so we have taken a lower estimate of the range of audibility at 10 km. Brandt et al. (2012c) found a significant deterrent effect on porpoises at ranges of at least 7.5 km for a Lofitech device, which greatly increases the previous known area of disturbance found by Olesiuk et al. (2002) to be at least 3.5 km. Neither study looked for effects beyond these maximum ranges. Again we have taken the lower figure, and set the range of deterrence at 3.5 km. By applying these figures to the locations of all Scottish Environment Protection Agency ( SEPA) licensed fish farms sites in Scotland, we estimate that the theoretical marine area of 'deterrence' is 3500 km ² and the area of audibility to harbour porpoise could be as high as 12600 km ² (figures 5 & 6) (assuming no shadowing of the signal by islands, as well as other assumptions). This represents around 4% and 15% respectively of the total inshore Scottish waters (<12 nm offshore).

It is important to note that these figures are not presented as estimates of the current extent of ADD audibility, but rather the potential extent, assuming that all SEPA licensed sites began using high-powered ADDs. In reality, many sites are inactive or fallow for at least parts of the year and only around half of active sites currently use ADDs. These figures also do not take into account the effect of bathymetry or the shadowing effect of landmasses, which would reduce these figures considerably. While these figures should be viewed as very rough estimates of potential maximum areas, they indicate a likely maximum percentage of Scottish coastal waters that could be ensonified, with the West Coast and Outer Isles most greatly affected. Further research could extend this concept by incorporating realistic usage patterns and propagation models in order to achieve a more reliable estimate of the likely marine area affected. Similarly, field data, such as that collected by the Hebridean Whale and Dolphin Trust ( HWDT) during routine monitoring cruises using towed hydrophones, could provide empirical data on the range at which devices are audible in 'real-world' noise conditions (Booth, 2010).

KG No.	Knowledge Gap
13	Over what maximum range are cetaceans likely to be impacted by ADDs?

Figure 5 Map of Potential Extent of ADD Audibility to Harbour Porpoise (Mainland and Hebrides)

Figure 6 Map of Potential Extent of ADD Audibility to Harbour Porpoise (Northern Isles)

2.2.3 Evidence of Efficacy

2.2.3.1 Current Knowledge

Strangely, far more scientific work has been done to assess the impacts of ADDs on non-target species than to quantify their efficacy for the purpose for which they were designed. While several studies have investigated effects of deterrent devices in contexts other than fish-farms, there is remarkably little published scientific evidence supporting their long-term use as effective pinniped deterrents in aquaculture.

While it may seem sensible to draw parallels from other operations that have been studied such as coastal salmon traps and salmon rivers, it is important to remember that from the predators' perspective, the context at a salmon farm may be very different and may therefore elicit a very different suite of behavioural responses. In particular, the motivation afforded by a cage full of large salmon, probably releasing auditory and olfactory cues, may dramatically increase the attractiveness of the site to the predator. Prior hunting success at a site ( e.g. before installation of anti-predator measures) could also influence the predator's choice. In addition, it is known that fish-farm sites often support an ancillary ecosystem which may include large numbers of wild fish, and anecdotal evidence from farmers suggests that these wild fish often play a role in attracting predators to the site.

Acoustic deterrents are referred to as a 'valuable tool for growers' in North America by NMFS (1996), but they present no evidence to support this claim apart from mentioning that the new high-powered (Airmar) systems were proving effective, and have become standard equipment for much of the industry in the Gulf of Maine. By contrast, Iwama et al. (1997), reviewed relevant literature and concluded that ADD effectiveness was highly variable among British Columbia aquaculture sites. They noted that any effect appeared to diminish with time, and that pinniped attacks continued to occur even when deterrents were present. They therefore recommended the phasing out and prohibition of acoustic deterrent devices, and this recommendation appears to have been adopted by the Canadian Department of Fisheries and Oceans ( DFO) for the British Columbia aquaculture industry, who are no longer issuing letters of authority required for installation of an ADD ( BC Pacific Salmon Forum, 2007).

2.2.3.2 Questionnaire Survey Studies

There have been several attempts to collect information on predator interactions using questionnaire surveys. This is a low cost method of consolidating information from a large number of geographically dispersed sites, allowing a broad perspective on the issue to be obtained. However, this technique has certain shortcomings, including the difficulty of exploring specific issues in detail, the potential for bias in the respondents to the surveys and the fact that data collected are often opinions rather than demonstrable facts. Questionnaire surveys are certainly a useful first step for exploring the problem and can generate testable hypotheses, but they should usually be followed up by directed research to collect and test real data.

Ross (1988) discussed the use of seal scarers when they were still very new to the industry; they were first trialled on Scottish fish farms in 1984-1985. Even at this early stage opinions appeared to have been mixed, with two operators reporting habituation within two weeks of use and two others reporting limited or no effect. Some operators apparently considered them still effective after several months in use, with one claiming continuing efficacy after two years of use (though ADDs were used here in conjunction with lethal removal, suggesting less than 100% efficacy). The reason for such variation was unclear, but hypotheses proposed included differences in motivation between individual seals, ADD usage patterns ( i.e. intermittent vs continuous usage) and factors specific to particular sites.

Rueggeberg and Booth (1989) surveyed British Columbia salmon farms and reported that five out of eight farms that had used them rated acoustic deterrents to be effective against seals, and one found them effective against sea lions. None of these devices, however, had been in use for longer than two months. Later studies in British Columbia found no grounds to support their use and ADD use in this area is now prohibited (Iwama et al., 1997).

Tillapaugh et al. (1993), summarised by Smith (1994), reported on the results of a 1991 questionnaire survey of 40 growers in British Columbia and concluded that, " in general, visual, auditory and sensory methods [of deterrence] were not effective".

Arnold (1992) investigated salmon farms in Shetland on behalf of Greenpeace UK, and stated that, " operators who have used seal scarers say that they can be mechanically unreliable and do not function for long as a deterrent due to habituation, and seals may even be attracted to the site". In a telephone and paper based questionnaire survey of Scottish salmon farm sites, Quick et al. (2002) found only 23% (21 of 92) of managers considered ADDs to be very effective with 6.5% reporting them to be completely ineffective. The majority of managers who responded felt that seal scarers were at least moderately effective.

Sepulveda and Oliva (2005) used questionnaires to assess the extent of South American sea lion predation at 48 salmon farm sites in Chile. Of the 16 sites that were using acoustic devices, 2 described them as 'efficient', 2 as 'moderately efficient' and 12 reported that they were 'inefficient'. They concluded that ADDs were reported to be ineffective in the long-term. The opinion of fish farmers seemed to be that they worked for 2-4 months, and then were no longer effective.

Nelson et al. (2006) surveyed a total of 97 Atlantic salmon farms in Maine, USA, and modelled the influence of a number of factors, including range to haulout and ADD use on depredation losses. They found that sites which utilised ADDs had a higher incidence of predation than those that did not. The authors concluded that their results showed ADDs to be ineffective, but this seems quite over-simplistic given their limited dataset. It may have been the case, for example, that only sites with particularly high levels of predation were using AHDs. Despite these unpromising findings, 50% of farm managers surveyed felt AHDs were 'fairly effective' and 6% reported they were 'completely effective'.

Northridge et al. (2010) conducted detailed on-site interviews with a sample of managers at salmon farms in Scotland. Three quarters (15 of 20) of sites where an opinion was expressed judged ADDs to have 'some preventative effect', while one quarter (5 of 20) said they had no beneficial effect.

The results of these surveys present a mixed picture. It is clear that ADDs do not provide a complete solution, but among operators and site managers there is significant support for their use. Within an area, opinions and experience of efficacy seem to be highly variable. In part this might reflect the lack of any formal experiments. Industrial practice tends to rely on perceived efficacy and there are few opportunities to risk a change in established operating procedures. In several cases there are indications that the efficacy of devices decreases with time, which could be an indication of habituation or learned strategies for avoidance or for controlling responses on the part of the seals. It is also generally true that few researchers have considered how effectiveness might be defined, because it is clear that even a marginal decrease in predation could be considered effective under some circumstances.

KG No.	Knowledge Gap
14	How can the effectiveness of ADDs be measured and compared, and what level of effectiveness is tolerable?

2.2.3.3 Research at Fish Farms

Several studies have made attempts to investigate the effectiveness of ADDs at fish farm sites. Most have relied on the opinions of the site managers (as in questionnaire surveys) but a few have monitored the level of predation after the introduction of a particular acoustic device. None, however, has involved a robustly designed, long-term experimental approach that might unequivocally determine the degree of effectiveness of a specific device. It is important to remember that effectiveness might, in this context, mean a modest reduction in the rate or severity of predation, as long as this reduction can be clearly demonstrated.

Pemberton and Shaughnessy (1993) described studies with two types of seal scarer (about which further information is unavailable) at fish farms in attempts to deter Australian fur seals. One operated at 28 kHz, the other at 10 kHz, and they tested each at fish-farms as well as haulout sites. The higher frequency scarer was tested at three farms over 6 months, during which 60 attacks were recorded. The lower frequency scarer trial ran for just two weeks during which three major attacks were recorded within five metres of the device. The authors noted that they were not able to determine whether the rate of attack was reduced by the ADDs due to experimental design, however, this level of predation was clearly unacceptable and these devices could not have been considered as useful management tools. Unfortunately they did not specify the make or source level of the devices tested, making their results impossible to generalise.

Several unpublished documents summarising field research by the ADD manufacturer Ace-Aquatec claim high levels of success at deterring depredation (Ace-Hopkins, 2002a; Ace-Hopkins, 2002b; Ace-Hopkins, 2002c; Ace-Hopkins, 2004; Ace-Hopkins, 2006). Ace-Hopkins (2002a) claimed 100% efficaciousness (in medical research efficaciousness indicates success in controlled experimental trials, as opposed to effectiveness which describes 'real-world' results), at sites where no previous ADDs had been used. A farm in one trial was losing around 50 smolts per week to depredation until the Ace-Aquatec ADD was installed, after which the author states that no further losses were reported, although he does not state details of how this was measured. At another site where an ADD had previously been ineffective the Ace-Aquatec ADD did not solve the problem. It is encouraging that this one manufacturer has conducted research and presented results openly and they are to be commended for doing this. These reports certainly indicate that useful data can be collected in a relatively straight-forward manner with a modest research effort assuming industry co-operation. Ideally, such studies should be longer, more rigorous, should incorporate better controls and have more extensive reporting. Of course, credibility would be improved if they were conducted by independent researchers not affiliated with an ADD manufacturer or the industry.

Stewardson and Cawthorn (2004) make reference to a trial conducted by the New Zealand King Salmon Company, to deter fur seals from aquaculture sites. A Poseidon T88 ADD was tested over a thirty day trial at two sites. They reported that, " results suggested that an ADD used in conjunction with other measures may, at least have a temporary effect in reducing seal attacks". Unfortunately, more detailed information on this study and the device used is not available.

The only attempt to conduct a controlled trial of which we are aware is reported by Vilata et al. (2010). They tested the effectiveness of an Airmar device against South American sea lions by comparing predation between two sites. The sites were both stocked with fish of the same size and in the three months prior to the installation of the ADD the biomass of salmon predated was not significantly different at the two sites. After three months of preliminary data had been collected an ADD was introduced at one of the sites. Over the subsequent three month experimental period, there were statistically lower levels of predation at the site with the ADD. They also found that the site with the ADD experienced significantly less predation than it had done during the same period in the previous year. This work shows that an experimental approach is possible and can provide useful results. The authors note that their sample size was small and that replicates are required before conclusions can be drawn. Furthermore, the experimental period was just three months, which may not be long enough to show effects of habituation. The authors also mention another adjacent site where the same model of ADD was installed at the same time and found to be "totally ineffective".

2.2.3.4 Playbacks at Haulouts

Designing and implementing experimental trials on the effectiveness of ADDs at operational sites is extremely challenging and would necessarily involve exposing caged salmon to predation risks that would probably be deemed unacceptable to site operators. For this reason attempts have been made to examine relevant literature from other comparable scenarios.

Acoustic playback studies at haulout sites have been used as one way to assess pinniped responses to particular signals. The main benefit of this type of experiment is that animals can be reliably found in useful numbers, and relatively large amounts of behavioural data can be collected over a short period of time.

A brief description of an early playback experiment by Pemberton and Shaughnessy (1993) reported that Australian and New Zealand fur seals were not deterred by either of two devices, one operating at 10 kHz, the other at 27 kHz. Ten trials were conducted with each device and the behaviour of the animals was noted. Seals reportedly continued with pre-playback behaviour, approached the device or raised their head from the water and looked toward the scarer.

The reaction of harbour seals to an Airmar ADD was investigated by Jacobs and Terhune (2002), who recorded seal behaviour using video while measuring distances from the sound source using an optical range finder. They conducted 16 treatments over 6 days, including 5 controls where transducers were placed into the water but no sound was played. No apparent difference in behaviour was found between treatments when the ADD was active and inactive, and no observable reactions (such as rapid swimming or hauling out) were noted. The closest sighting of a seal while the device was active was 43m. They also tested whether an active ADD would prevent movement of seals through a channel by counting the number of animals at a haul-out which could only be accessed by passing within approximately 600m of the device. No effect was observed between treatments and seals were observed as close as 44m from the sound source.

Stewardson and Cawthorn (2004) describe the results of an experiment in New Zealand, where fur seals at coastal haulout site were exposed to a device manufactured in Sweden by Kemers Maskin AB. The device had a source level of 200 - 210 dB (re 1 µPa - RMS or peak not stated), with a frequency of 10 kHz. Animals were classified into size classes in order to look for differences between the responses of age groups. Small and medium sized fur seals made "convulsive changes of direction and porpoised [rapid surfacing during directed swimming behaviour] rapidly from the sound source". In contrast however, large adult males were reported to "initially show indifference before making a positive response to the sound of the ADD."

Götz (2008) tested the response of grey seals to eight different sounds (including white noise, a 500 kHz sine wave and those produced by currently available seal scarers - Lofitech, Airmar, Ace-Aquatec and Terecos), at haulout sites in the Tay Estuary, Scotland. Playbacks were done from the stern of an anchored boat, using visual counts on animals within five distance bins (0 - 20, 20 - 40, 40 - 60, 60 - 80 and 80 - 100 m) to compare between pre-sound, sound and post-sound treatments. During control experiments, when no sound was played, no significant changes in animal numbers were shown. For all tested sound types (except one - the Terecos sound type), there was a significant decrease in the number of animals in at least one of the distance ranges (repeated measures ANOVAs all p < 0.05). To test for evidence of habituation within each day, they counted the total number of seals within 60m for each playback. There was no significant relationship between playback number and count of seals, and therefore no evidence of decreasing deterrence effect or habituation. They also did not find any evidence of animals being attracted to the sound source, as was reported by Pemberton and Shaughnessy (1993).

These studies show that playback experiments in the wild may be a useful tool for assessing the relative aversiveness of particular signal types. However, the context is different from that at a fish farm in several important ways. For instance, depredating animals may be highly motivated to feed on an abundant food source and ADDs are often in operation for much more extensive periods of time which may have an effect on the way the signal is perceived. In addition, avoidance tends to be measured at quite large ranges whereas depredation reduction could result from animals being deterred by just a few metres in order to reach the nets. We therefore question how appropriate or useful such studies are for informing depredation management strategies.

KG No.	Knowledge Gap
15	The effectiveness of ADDs in reducing seal depredation to stocked fish remains unclear. An experimental approach to address this fundamental uncertainty is difficult for economic and fish welfare reasons.
16	Effect of motivational state and context in mediating and modifying aversive response to ADDs.

2.3 Alternative Approaches to Managing Seal Interactions at Fish Farms

2.3.1 Containment

Northridge et al. (2010) reported that fish farm operators had suggested that problems with seal depredation have generally improved over the past decade or more. It is clear that this improvement is not the result of any increase in the overall proportion of sites using ADDs, which has remained around one half since the study by Quick et al. (2002). Most respondents indicated that improved containment and better husbandry have been the primary drivers behind the improvement. Without access to detailed industry data it is impossible to be sure how effective such measures have actually been in reducing seal depredation, but Scottish Government figures do at least suggest a decrease in the numbers of fish that are reported to have escaped due to breaches in cages over the past ten years (Northridge et al., 2013).

The particular measures that may have been responsible for a reduction in seal depredation at Scottish salmon farm sites also remain unclear, but several trends were noted by Northridge et al. (2012). Nets have for example generally increased in size over the past ten years or more, and salmon stocking density has been reduced. It has been suggested that high stocking densities may have made depredation easier for seals, and that lowering fish densities (for welfare and for improved productivity) may therefore have helped reduce depredation. Larger nets may also have made access to fish within nets more difficult for seals if fewer fish are to be found close to the net perimeters.

KG No.	Knowledge Gap
17	How does stocking density influence seal behaviour and depredation rate?
18	Salmon behaviour within nets and in response to depredation is poorly documented, particularly at night.

It should be stressed that at present we have very limited information on how seals actually take fish from inside nets without actually breaching the containment wall of the net pen itself. Net breaches are relatively rare, and by far the most common means of depredation appears to be grabbing fish through the meshes of the net wall or floor of a fish pen and sucking flesh through the netting. Exactly how this is usually done is unknown, although there are a few anecdotal accounts (see section 2.1.3 for more detail).

The second change that has occurred in Scottish fish farms has been the gradual increase in weighting used to maintain net shape and prevent deformation by tidal currents through improved net tensioning. Surprisingly little research has been conducted in this field, and we were only able to identify one field study that has examined net deformation in salmon cages under different weighting regimes and tidal current systems. Lader et al. (2008) compared the net deformation in two sites with different weighting systems and with exposure to different current regimes in Norway and in the Faroe Islands. At one site with square pens and with current speeds of 0.13 ms ^-1 (0.25 knots) there was an estimated 20% reduction in net volume at peak tidal flow, while in the other a 40% reduction in circular pen volume was measured when current speeds reached around 0.35 ms ^-1 (0.68 knots). The square pens were fitted with weights of 2 x 600kg, 400kg and 300kg at each corner and 2 x 125 kg weights on each of two sides and 2 x 80kg weight on each of the other two sides. This is more than is normally found at Scottish square pen sites. The circular pen was fitted with a sinker tube (single ring weight) of 1700kg. Neither of the current speeds in this study was particularly fast compared with those experienced in Scotland. We have been told that individual weights used on Scottish net pens have increased from 20kg to 80kg or more over the past ten years, and, during visits, we have also observed some extreme net displacement at Scottish fish farm sites at peak tidal flows despite this increase in weighting.

Net displacement decreases the volume of the net (see Figure 7 below from Northridge et al., 2013). This not only increases the effective stocking density but also likely results in net deformations that may make it easier for a seal to attack salmon through the meshes, especially where pockets are formed (see also Section 2.1.3).

There has also been a trend in Scottish farms towards the use of circular pens which are usually weighted with a circular basal weight of 1.5 tonnes or more. This system is likely to maintain net structure more firmly through the tidal cycle and thereby reduce the chances of net pockets or other deformations arising. Preliminary evidence suggests that successful depredation is more limited in circular net pens than in the older square pen designs (Northridge et al., 2013; Thistle Environmental Partnership, 2010).

It is therefore likely that increased weighting and better net tensioning could have played a significant role in diminishing seal depredation, and it remains possible that further improvements in this respect may lead to further declines in seal depredation.

KG No.	Knowledge Gap
19	How does net tensioning affect the ability of seals to remove fish?

Figure 7 Net displacement as a result of tidal flow

Other measures that have been cited as possible means of minimising seal depredation include the use of anti-predator nets, the use of seal blinds and prompt removal of salmon 'morts' (dead fish) from cages. There is no empirical evidence to show how effective these measures may be. There is a widespread belief within industry that dead salmon attract seals (through smell), make easy pickings and also encourage depredation more generally. The extent to which this is actually true is not known.

Seal blinds are relatively small sections ( e.g. 2m x 2m) of small meshed netting that are sewn into the base of cages, at the deepest point or vertex of the cone or pyramid that forms the net base, where dead fish are likely to accumulate. The idea is to prevent seals from seeing dead fish and make the net stiffer and more difficult to deform. Some nets use seal blinds made from stiff plastic mesh rather than netting. The efficacy of this strategy has not been evaluated objectively, but the fact that seal blinds are not uncommon design features of net cages suggests that they must be at least be perceived as partially effective.

KG No.	Knowledge Gap
20	How important are dead fish (morts), and their removal or concealment in motivating or preventing seal depredation?

Anti-predator nets are additional, usually large mesh, nets that surround each pen within a farm site. They may be deployed as curtains from the outer edge of the walkway around each pen falling to the seabed, or alternatively may be rigged as box nets to surround the sides and base of a net, or may even surround the entire site. Although Hawkins (1985) found they were used at 88% of Scottish farms, such nets are now rarely used in Scotland because they are difficult to manage, are liable to foul the propellers of tending boats, add to mooring problems caused during bad weather, impede the flow of water through the pens, can catch large numbers of sea birds and other marine wildlife including seals and cetaceans, and are anecdotally not thought to be very effective, as seals are often able to penetrate them anyway. Northridge et al. (2010) found anti-predator nets in use at just one of 136 sites. Despite these domestic reservations, such nets are widely used in other countries (Canada, Chile and Australia), and it is as yet unclear how or why they appear to be effective in these countries.

KG No.	Knowledge Gap
21	How are anti-predator nets utilised internationally to avoid common problems experienced in Scotland?

One possible reason for the continued international use of predator nets is that there has been some degree of dedicated research into the most effective net types and configurations elsewhere. These research efforts appear to have made some progress toward limiting the problems commonly associated with anti-predator netting in Scotland. Schotte and Pemberton (2002) report that anti-predator nets were often used to protect Tasmanian salmon farm sites from New Zealand fur seal ( Arctocephalus forsteri) and Australian fur seal ( Arctocephalus pusillus) once fish were over 300g. One common problem encountered with their use was that even under modest degrees of current the two layers of netting (anti-predator and growth netting) can come together. To overcome this, tensioning weight must be distributed between the two nets, and various practical methods of doing so were described. The use of neutrally buoyant steel pipes, individually cut to length and known as 'separation sticks', was also found to increase the distance between the nets and prevent them from coming together. This method was particularly recommended for use between the base of the growth and anti-predator netting, where seals are thought to push nets upward in order to scavenge dead fish.

Schotte and Pemberton (2002) also developed a scale model to investigate the effects of water flow on different configurations of tensioning weight. They found that net shape ( e.g. the degree of tapering between the waterline and net base) was an important factor but was very difficult to perfect in practice, due in part to shrinkage after deployment. Based on this work they recommended that a minimum of 20% of available buoyancy should be used for net tensioning, distributed as appropriate between the two layers of netting. For a 120m circular pen (the largest used in Scotland) this would mean the use of 2.4 tonnes of weight. They also recommend a minimum spacing between the growth and anti-predator netting of 2m. This distance is practically achievable when suspended from the outside of steel walkways, but for plastic circular walkways would probably involve the addition of an extra ring (which would also increase the amount of buoyancy available for tensioning weight).

In conclusion, there is some evidence that improvements in containment may have helped reduce incidents of seal depredation at Scottish salmon farms. Various individual measures have been proposed as being responsible, but as yet it has not been possible to separate out the effects of each of these gradual improvements. It is entirely possible that further improvements (for example in net tensioning) may further reduce seal depredation, but there are no ongoing studies to assess these issues at present.

2.3.2 Lethal Removal

The lethal removal of predators (as an act of last resort) is an emotive issue and raises conservation, ethical and welfare concerns. Shooting seals at fish farms has been carried out since the industry was first established in Scotland, but has become more strictly controlled in recent times. Hawkins (1985) found that 20 of 41 sites used shooting to protect stock, and this remained true until at least 2001, when Quick et al. (2002) again found shooting at approximately 50% of sites - though the total number of animals was not given.

The Scottish Salmon Growers Association, as early as their 1990 code of practice, adopted the policy that lethal removal should only be used after all reasonable attempts have been made to exclude seals with non-lethal methods, and this has been adopted in the most recent code of good practice (2010):-

"Seals should not be shot during their close seasons (common seals 1 June to 31 Aug; grey seals 1 Sep to 31 Dec) unless all reasonable attempts have been made to apply exclusion measures, these have proved to be ineffective, and there is a significant risk of damage to fish and fish farms."

A licensing scheme covering lethal removal has now been introduced under the Marine (Scotland) Act 2010 which is providing, for the first time, data on the number of seals lethally removed from fish farm sites.

The logic of selective lethal removal as a management action depends largely upon there being a small number of "rogue" animals causing the majority of the damage at a site, and on the ability of a marksman to reliably identify and remove these individuals. It is typically the case that there will be many seals at a fish farm site for extended periods with no serious depredation incidents (Northridge et al., 2010). Many fish farm managers believe that problem animals exist and can be identified and removed, but there is no independent evidence to support this assertion. The main indication that rogue seals exist and can be removed in this way is that farmers often report at least temporary relief from seal depredation after one or more seals have been removed by shooting. This information is largely anecdotal in the fish farm context however. One scientific discussion of lethal removal can be found in an article by Pemberton and Shaughnessy (1993), who dismissed shooting as ineffective, although it is not clear from their work whether they are referring to lethal removal, or simply its use as a deterrent. They refer to one seal which was 'shot at' at least 30 times and did not appear to leave the area. An alternative explanation could be that shooting frightens the remaining seals rather than removing a particular culprit.

KG No.	Knowledge Gap
22	The "rogue seal" hypothesis, and the rate at which removed seals are replaced is currently unclear.

Despite a lack of evidence about the efficacy of lethal removal as a management technique, this is certainly an area where objective data could be relatively easily collected. For example, such data might include: the recent history of depredation incidents at a farm site; the number, species and identity of seals seen at sites during these incidents (using photo-id); the number, species and identity of seals removed and details of any subsequent depredation. If carcasses of shot seals were recovered then their identity and recent diet could be determined, which would be highly valuable in determining efficacy of lethal removal methods. These carcasses would also be a uniquely valuable source of data for other applied research, including population studies.

It is a condition of the licence agreement under the Marine (Scotland) Act 2010, that, "The licensee must take all reasonable steps to recover the carcases of shot seals but only when it is safe to do so. … Even a carcase which has been in the water for several days should be retrieved wherever possible. " It is however, difficult to recover carcasses in practice, and there is a logistical limit on the ability of the Scottish Rural College ( SRUC) to recover such animals. Anecdotal discussion with experienced marksmen suggests that approximately half of shot seals will immediately sink. Thus far, very few carcasses have been recovered from several hundred shot at fish farms annually.

KG No.	Knowledge Gap
23	How can recovery of seal carcasses be improved?

Animal welfare concerns relating to lethal removal focus on the possibility that seals may not be killed instantly and could be injured and suffer serious pain as a result. Licensing conditions reduce the possibility of this occurring by specifying the type and calibre of weapon that should be used and the experience of marksmen employed. We believe that larger fish farm companies are increasingly using a small number of experienced marksmen to carry out lethal removals, which will be helpful. Information on the reliability with which seals are shot "cleanly" could be collected by independent observers and by post mortem of carcasses of shot animals (as is currently the case for the few animals recovered).

Conservation concerns will arise if lethal removals at salmon farms are contributing to a level of mortality which is unsustainable or prevents recovery of depleted populations. In Scottish waters, these concerns will be higher for harbour seals, whose populations have declined substantially in some areas in recent years, rather than for grey seals ( SCOS, 2011). The Scottish Licensing scheme allocates "quotas" to the industry which should ensure that total removals do not exceed Potential Biological Removal ( PBR). This system requires that the species of seals being removed is determined reliably and that certain demographic components of the population are not over-represented. This is another area where observer data and recovery of corpses could provide useful data.

KG No.	Knowledge Gap
24	How can information about the demographic parameters of seals shot be improved?

Seals are an important part of Scotland's ecotourism resource and there is a conflict of interest when lethal removals take place in areas where seals are an important source of income for ecotourism operators. Lethal removals could potentially affect these activities by reducing numbers, displacing animals from local haulouts and making them less approachable, and possibly discouraging tourists.

A further factor which must be considered is the wider societal concern over lethal removal of wild animals (particularly iconic mammals), especially without a clear and proven link to effective management. This means of management could damage the public perception of individual salmon farming companies, the retailers that sell their products and the Scottish aquaculture industry as a whole. A number of campaigning organisations are opposed to seal shooting even under licence.

It is therefore important to establish as clearly as possible whether lethal management is effective in minimising damage to fish farms.

KG No.	Knowledge Gap
25	It is unclear to what extent lethal removal is effective in minimising damage. No studies have looked at how depredation rate is affected by lethal removal.

2.3.3 Conditioned Taste Aversion

2.3.3.1 Conditioned Taste Aversion and its Application for Controlling Terrestrial Predators

Conditioned Taste Aversion ( CTA) is a process by which an animal "learns" to avoid food which has made it ill in the past. Once an animal has been made ill by eating poisoned or tainted food it will usually exhibit disgust and may vomit when it encounters that food again. The ability to avoid ingesting poisonous or harmful substances is of such fundamental survival importance that CTA is found in all animals, from humans to sea anemones. This involves a specific form of learning mediated by dedicated neural pathways operating within the more 'primitive' parts of the nervous system and resistant to the influence of higher cognitive processes. Typically, this aversion is learnt after a single trial and persists for months or years. Conditioned taste aversion is also known as the "Garcia Effect" after John Garcia, an American psychologist who discovered that if an animal was made ill by some other mechanism, for example by exposure to radiation, it would develop a CTA response to a novel though harmless food stuff it had ingested during a time period before the sickness was induced. Thus, animals can "learn" an aversion to a food type even though it might have been made ill by some other agent.

Wildlife managers in other industries have made use of this phenomenon to develop non-lethal methods to reduce depredation by predators. The process usually involves lacing bait made from the flesh of the animal to be protected with an emetic (a substance that will induce vomiting) which itself is not detectable by taste. For example, in an early series of experiments (Gustavson, 1977; Gustavson et al., 1974; Gustavson et al., 1976) coyotes and wolves were fed minced sheep flesh laced with lithium chloride (an emetic) and wrapped in sheep wool (to provide cues similar to those a wolf would experience when attacking its prey). The predators became nauseous after eating the LiCl laced bait and after recovery they showed a marked reluctance to attack the prey they had been conditioned against. Gustavson and colleagues went on to show that wolf and coyote depredation of domestic sheep could be greatly reduced in this manner ( e.g. Gustavson, 1982; Gustavson et al., 1982).

There have been many attempts to control predation of terrestrial animals using CTA but not all have been successful ( e.g. Conover, 1989; Conover and Kessler, 1994). Some of the failures may be due to unrealistic presentation of the baits and lack of attention to detail.

Several particular aspects of CTA are important from the perspective of its use in predator control (Cowan et al., 2000):

• CTA is a form of learning that takes place very quickly, usually after only a single exposure, yet can be long lasting, persisting for months or years.
• The neural processes underpinning CTA take place within the oldest and most primitive part of the brain: the hind brain. Inputs can arrive from receptors in the gut as nerve signals transmitted via the vagus, part of the sympathetic nervous system, involved in regulating the bodies' internal activities and state. Chemicals in the blood stream may also stimulate areas of the hind brain directly. Thus, this primitive form of learning is subconscious and deep-seated, in fact it is impervious to influences from higher levels of the nervous system and the conscious mind.
• During the CTA process, an aversive association is established with substances that had been ingested within a time window, typically of one to six hours, before the onset of nausea. This interval matches the time that it would normally take for a toxic substance to be ingested, for digestion to begin and for it to either trigger toxin receptors in the stomach or for chemicals to enter the blood stream and have a direct effect in the hind brain.
• The taste, smell and flavour associated with the "suspect" food are the most readily and strongly conditioned cues. However, associations can also be made with other triggers, including visual cues, although this may require repeated exposures. CTA seems to be most easily established if the food stuff is novel but aversion can also be established for previously encountered foods.
• CTA is not associated with a particular location and the aversion applies to the food type wherever and in whatever context it is encountered.
• CTA can be very specific. For example, aversion will usually apply to one particular prey species. This allows a predator to avoid poisons while excluding as little from its normal diet as possible.
• CTA conditioned animals of different species have been described as behaving in very similar ways when exposed to the conditioned food. For example, when mammals such as coyotes and wolves taste or smell the conditioned food stuff, they shake their heads from side, wretch, urinate and move away.

2.3.3.2 CTA to Protect Salmon from Pinnipeds

There have been several attempts to use CTA to curb the predatory behaviour of pinnipeds and in particular to reduce predation on both wild and farmed salmon.

Lithium chloride is the most straight-forward emetic but its use has proven problematic with some terrestrial predators because they are able to sense its slightly salty taste. One might conjecture that this is unlikely to be a concern for a seal feeding in seawater.

The most complete and best controlled trials, reported by Kuljis (1984), were carried out with captive Californian sea lions ( Zalophus californianus) in the USA. Four male yearling sea lions were fed twice a day on a diet of herring and mackerel, with the fish species being alternated between feeding sessions. Both species of fish were highly preferred food and the sea lions always ate all the fish presented to them. After a 21 day pre-test period, two of the animals were fed mackerel laced with lithium chloride capsules. The other two seals were fed unadulterated mackerel. Within half an hour, both of the treated sea lions began vomiting and this continued for around 20 minutes. After an hour they seemed to recover fully and returned to their normal behaviour. The next feeding trial was of herring and all animals ate all of the fish presented to them. On the next mackerel feed however, the first fish offered was taken and then dropped. Subsequently, one of the treated sea lions refused to eat any mackerel for the next 19 days. The other treated sea lion returned to eating mackerel after 3 days. This sea lion was then given a second dose of lithium chloride after which its mackerel consumption was virtually eliminated. After 15 days, when some recovery in mackerel consumption had taken place, all four seals were dosed with LiCl at a lower concentration. All four then ceased taking mackerel for the next two days, after which the trial was terminated for unexplained reasons.

The treated animals were not offered any alternative source of food, so during the trials they were living on half the ration they had been used to before the trials began yet continued to refuse the mackerel offered them. The health of the subjects was carefully monitored through the trials and regular blood samples were taken. Blood parameters were within the normal range. In a subsequent set of trials, conducted to investigate potential health issues, herring were laced with emetics and the test subjects ceased feeding on them but continued to feed on mackerel (Costa, 1986). More detailed blood tests during these experiments failed to reveal any significant changes. Field trials with wild pinnipeds were later conducted as part of attempts to reduce sea lion predation of wild salmonids migrating through locks in rivers in Oregon. Freshly killed steelhead trout were suspended on a line in the water near the locks. Once sea lions were taking these unadulterated baits routinely, trout laced with LiCl were substituted. Two sea lions that could be identified in the field took these treated trout, they then disappeared from their normal foraging locations but returned within 2 hours. These animals would no longer take the tethered fish but it was not clear whether this apparent aversion had been extended to an avoidance of live fish too.

Conditioned taste aversion trials have been carried out at salmon farms in Tasmania where the main pinniped predators are large male Australian fur seals ( Artocephalus pusillus doriferus). Pemberton and Shaughnessy (1993) summarised the results of 26 trials each initiated by seal presence at a fish farm site. Trials started with a period during which unadulterated whole salmon baits were presented. These were then replaced by LiCl treated baits. On 21 of 26 occasions fur seals took the treated bait and on four of these, seals were seen convulsing and vomiting by either researchers or fish farm workers and subsequently left the fish farms. These preliminary trials were considered successful and Pemberton and Shaughnessy (1993) recommended that the technique should be further developed. However, in spite of this encouraging start, no further work seems to have taken place.

No CTA trials have been attempted at Scottish salmon farms, nor indeed with any phocid seals. However, in 1988 some ad hoc tests of repellents were undertaken in Loch Sunart. Salmon of large smolt size were laced with chilli and curry powder and suspended on lines from fish farm cages. Seals, believed to be grey seals, took these baits and although seals continued to take bait over the period of the trials and were not repelled, seal problems at the farm reportedly ceased during the trial. Unfortunately, the experiment was not taken any further. It is important to emphasise that this was a trial of a repellent and not of conditioned taste aversion. In conditioned taste aversion it is important that the predator should not detect any difference, especially in taste, between the bait and the prey species generally. In this way, protection should be generalised. This brief trial illustrates that wild Scottish seals will readily take whole dead salmon bait presented at a fish farm.

2.3.3.3 Research and Development

While the research projects summarised above provide strong grounds for optimism, it is clear that should CTA be developed to a state where it can be used as a routine method for controlling seal depredation additional trials and research would need to be carried out in several areas.

KG No.	Knowledge Gap
26	Which emetics are most effective and what are the minimum doses required for CTA?
27	Are there any harmful physiological effects on seals treated with CTA, and if so, how can they be minimised. Is CTA sufficiently specific to salmon to leave the seals' normal diet of wild fish unaffected?
28	Are there any environmental effects of CTA?
29	How can baits for CTA best be prepared and presented to wild seals at salmon farms?
30	What patterns of "treatment" are most effective? Should baits be presented routinely or only when problems become evident?
31	Are there seasonal difference in when and how CTA should be use? Should there be "closed seasons"?

Answering these questions should be relatively straight-forward and could be achieved using routine research methods applied to both captive animals ( KG numbers 23, 24 and 28) and trials with wild seals at fish farm sites ( KG numbers 25, 26, 27 and 28).

2.3.3.4 Conclusion

The work on conditioned taste aversion for controlling depredation by pinnipeds is, taken as a whole, rather encouraging, especially when the rather meagre research effort expended is taken into account. Given this, it is surprising that there has so far been little appetite for developing this methodology for use in Scotland. There seem to be concerns within the industry of a public perception that seals being made ill by eating adulterated salmon could indicate that farmed salmon is itself tainted. Good public awareness initiatives linked to publicity of the real welfare and environmental issues that CTA could circumvent, should diffuse this perceived issue.

2.3.4 Electric Fields

Protecting caged fish from attacks by seals usually relies on producing an avoidance reaction at sufficient distance to dissuade seals from approaching nets. Such methods have generally relied on acoustic deterrents of some form, as described above. An alternative approach may be to produce an avoidance reaction over a very small range to prevent the seals from pushing against (or biting through) the net itself. This approach has the advantage that it produces the minimum possible change in the behaviour and distribution of predators.

Studies in fresh water appear to indicate that seals may respond to electric fields at strengths significantly lower than those which cause behavioural responses in salmonid fish. Forrest et al. (2009) have shown that seals can be excluded from some freshwater systems by the use of electric fields. A useable freshwater deterrence system has been developed and tested in the US and Canada on both captive and wild, free ranging seals and sea lions with promising results (Forrest et al., 2009).

The North American installations were operated in low conductivity water (25 to 250 µS/cm [micro Siemens per centimetre]) yet still required substantial (kilowatt) power supplies. Unfortunately, replication of the same approach in seawater would require large amounts of electrical power. This is because the electric field required to stimulate an organism is likely to reduce only slowly with increasing water conductivity (Lines and Kestin, 2004) while the power required to sustain an electric field in water increases in direct proportion to the water conductivity and water volume. Sea water has a conductivity of 45000-60000 µS/cm - far more conductive than the water used by Forrest et al. (2009) so the power requirement to replicate their setups in sea water would be between 200 and 1000 times greater than that used in the original trials.

The very large electrical currents needed to sustain an electrical field in sea water means that there is little chance that this fresh water approach would be practical in the marine environment. It should however be possible to produce local electric fields within a few centimetres of a net wall with deterrent capabilities that could eventually be developed into a system for preventing direct attacks on net cages.

Milne et al. (2012) describe a preliminary series of behavioural response trials to assess the effectiveness of pulsed low voltage electric fields to deter trained captive seals from entering a feeding station. Seals were trained to enter an outer tube and maintain station with their noses close to and usually in contact with a Perspex feeding port within which they could see a small food reward. A Perspex window above the outer tube allowed the observation of any behavioural response. Seals were required to hold station for approximately 3 seconds before the port was opened to allow access to the food reward.

During trials electrodes were placed at either side of the entrance to the feeding port so that seals needed to position their heads in between the electrodes in order to gain access to food items. A single test comprised a series of 5 control trials with the device off followed by 5 trials with the electrodes energised. Trials were carried out at a range of voltages 12, 18, 24, 30 and 36 V and for each voltage at a series of different pulse durations of 10, 20, 50, 100, 200, 500 and 1000 micro-seconds and at pulse rates of 10, 50 and 100 Hz. Trials were carried out with juvenile grey seals, juvenile harbour seals and an adult male harbour seal. To avoid accidental exposure to painful stimuli the trials were not carried out in a randomised sequence. Instead seals were exposed to gradually increasing field strength, signal durations and pulse rates, starting at levels that were expected to be undetectable.

Results clearly show that both seal species are able to detect low voltage pulsed electric fields. There is also strong evidence that the level of response is affected by changes in the combination of pulse duration, voltage and pulse repetition rate. None of the seals showed any signs of detecting or responding to the electric field at short pulse durations of 10 or 20 µs, but all seals showed clear aversive responses when exposed to longer pulse durations and higher voltages. This was especially apparent at high pulse rates (100 Hz) where there was clear evidence of avoidance with seals refusing to push through the field to access food items.

These reactions were transitory in that all seals continued to use the feeding system after they had refused to enter the feeding tube. In all cases seals returned to use the feeder during the same experimental session. After refusal there were indications that seals were more cautious in their approach on the next trial, but thereafter they returned to the device as usual. The preliminary results showed no clear sign of either sensitisation or habituation, but the trials conducted so far have had limited power to detect either.

In all cases the extent of the electric field or at least the extent of the effect zone was small. At the higher voltages, pulse lengths and pulse rates the seals pulled back from the entrance to the feeder where the field was most intense. However, even at the highest settings the seals usually stayed inside the outer tube and held station within 30 cm of the feeding port.

These results are encouraging and suggest that low voltage, low duty cycle, pulsed electric fields might prevent seals from pushing against a net. However, the results are preliminary and further trials will be required to assess the effectiveness of different electric field patterns with reduced energy input in order to approach a useable net protection method.

KG No.	Knowledge Gap
32	Behavioural aspects of electrical deterrence: how will behaviour be modified by context and motivation?
33	Practical aspects of electrical deterrence: engineering solutions are lacking and will be required before deployment in real-world applications can be feasible.

2.3.5 Trapping for Translocation, Conditioning or Lethal Removal

Capturing problem pinnipeds in traps has been a routine management method applied for protection of migrating salmonids at locks in rivers in the USA (Brown et al., 2008). Capture and removal has also been used at some salmon aquaculture sites in Tasmania ( e.g. Robinson et al., 2008b) and in specially adapted salmon traps in the Baltic Sea (Lehtonen and Suuronen, 2010). Designs for a floating trap for this purpose are available in a patent (Sandlofer, 1989) and Brown et al. (2008) describes the use of a significant number similar devices at locks on rivers. Once seals were captured at aquaculture sites, farm managers would be able to make an informed assessment of whether or not the captured seal was a "problem animal". Certainly, they should be much better able to do this reliably than could a marksman. Stomach lavage or scat analysis might reveal evidence that salmon had been eaten, for example. There might also be an immediate indication of whether depredation at the site had ceased. Confirmed "problem" seals could be translocated and released in another location, conditioned to avoid salmon or euthanized.

Capture and translocation has been used extensively in the Tasmanian aquaculture industry, where 4517 translocations of Australian and New Zealand fur seals were made between 1990 and 2005 (Robinson et al., 2008b). This was done as a commercial enterprise with some cost absorbed by the Government Department of Primary Industries, Water and Environment. Seals were moved approximately 400km along the coast before being released, but a high number of animals found their way back to the original sites. In 2001, 38% of those animals captured had previously been translocated. The average interval between recaptures was calculated as 38 days for NZ fur seals and 30 days for Australian fur seals, showing that a short-term respite from attacks can be achieved (Robinson et al., 2008a).

Where this approach has been applied it has often provided only short term relief due to animals returning (Anon., 2002a). In Scotland, there is such a high density of aquaculture sites on the west coast and Western and Northern Isles that it would seem likely that a "problem animal" might also soon find an alternative farm and translocation might be a case of moving a problem rather than solving it. Moving animals from the west coast to the east (where salmon farms are absent) could lead to concerns amongst wild salmon fishermen and sportsmen who also experience seal depredation. Brown et al. (2008) reported that in the US some trapped animals are provided to captive facilities. If the number of seals trapped was to match the numbers currently shot in Scotland (several hundred per annum) it is unlikely that UK or even European captive facilities would have the capacity to accommodate the number of animals involved.

Another potential alternative would be to condition the captured animals to avoid salmon. This might be done using conditioned taste aversion for example (see section 2.3.3). If the captured animals had to be destroyed (as a last resort) then at least they could be killed humanely after a reasonable and justifiable assessment had been made that they were indeed the culprit animal. Traps would need to be specially built and tended, so this would certainly involve expenditure. However, seal depredation can result in very substantial losses and this approach might not necessarily be more expensive than other options, such as expensive ADDs and the cost of bringing in a specialist marksman. It is a method that could be applied as soon as a seal depredation problem started.

KG No.	Knowledge Gap
34	What legal and ethical restrictions would affect the use of trapping for translocation, conditioning or lethal removal?
35	What would the monetary cost of implementing such a trapping system be?