Aquaculture - use and efficacy of Acoustic Deterrent Devices (ADDs): report

5 Future research priorities

5.1 ADD extent of use

From the data sources considered in this project, information presented in the seal licence survey to support applications have been the most consistent source of information on ADD use over time. These surveys were not systematic or rigorously controlled, but provided a centralised record of broadscale ADD use over several years. With the changes to the seal licensing regime that came into force on the 1st of February 2021, such information will no longer be collected through this approach. Contributions to this study from finfish producers and ADD manufacturers were variable in content and dependent on their own degree of record keeping and willingness to share information. Information from accreditation schemes were limited to those farms and producers signed up to the scheme. Furthermore, data collected by these schemes is not standardised or necessarily accessible for research purposes.

Systematic record keeping and centralised data collation is essential to ensure that ADD use can be monitored, assessed and managed effectively. This should include all information relevant to understanding the sound output from ADDs (the ADD model and operating mode, duty cycle and number of transducers) relating to each finfish farm site and should be collected at an appropriate temporal scale to assist in answering key research questions. At least daily record-keeping is recommended. A centralised online system with automated data logging of ADD status might be a suitable solution for collecting such data at a very high temporal resolution.

5.2 Alternative non-lethal measures

Alternative measures for managing depredation were considered in detail under this project, and results have been included in an associated project report (Thompson et al., 2021). Research recommendations relating to alternative measures considered as part of that report include the use of novel physical barriers and netting materials, conditioned taste aversion, non-lethal removal of predators (trapping) and the development of methods for real-time predator detection.

5.3 Impacts on non-target species

The use of ADDs inevitably increases the level of anthropogenic underwater noise, which is recognised as a potential chronic stressor of marine mammals (OSPAR, 2014). However, the ecological significance of this increase is not easily defined. This section gathers relevant scientific literature on the use of ADDs and the potential for impacts on marine mammals.

There are three key ways in which non-target species may be impacted by the use of ADDs; displacement, habitat degradation and auditory injury (e.g. hearing threshold shifts). Failure to consider all potential pathways of effects is likely to lead to underestimation of true cumulative impacts (Williams et al., 2020). Each of these pathways are considered in turn.

Research is recommended on the potential impacts of ADDs on each of the marine mammal hearing groups defined by Southall et al. (2019), low frequency, high frequency and very high frequency, all of which are present in Scottish waters. Further understanding of the population size and trajectory, as well as the ecology of the species in these groups, would assist in defining and understanding the potential significance of any impacts. The minke whale is a low frequency species which is relatively abundant around Scotland, and has been successfully studied in an ADD exposure experiment (McGarry et al., 2017). Bottlenose dolphins and harbour porpoise are relatively abundant around Scotland and are suitable candidates for studying the potential impact of ADDs on high frequency and very high frequency hearing groups, respectively.

There are three pathways of potential impacts on non-target species which require investigation: the potential for disturbance and displacement, the risk of habitat degradation (e.g. through auditory masking), and the likelihood of auditory injury (PTS). Habitat exclusion and the likelihood of auditory injury are closely linked, because the risk of auditory injury is dependent on cumulative exposure levels to ADD noise, which is influenced by changes in animal distribution in response to noise. Potential approaches to understanding these issues are considered below, followed by a discussion of the need for development of population-level approaches, which applies across all the pathways of potential impacts.

Displacement and disturbance

The question of whether ADDs installed at Scottish finfish farms can disturb marine mammals in an ecologically significant way is complex to answer, and research has so far not fully answered the question.

There is evidence that ADD noise is widespread in the Scottish marine environment. An increasing trend in ADD detections within the Inner Hebrides and the Minches SAC was found from 2006 to 2016 (Findlay et al., 2018), with a peak in 2013 (12.6% of sampled locations) suggesting that a significant proportion of the monitored habitat is acoustically affected by their use. Evidence from this report shows that the number of farms using ADDs has increased since 2016 to 2019, and so the proportion of impacted habitat may also have increased. Furthermore, the audible range of ADDs is likely to be many kilometres or tens of kilometres, and will vary depending on signal and local propagation characteristics. Simple geometric models of sound propagation have been found to be inadequate for predicting complex sound fields in near-shore environments (Shapiro et al., 2009).

There is currently no established methodology in the UK for assessment of marine mammal disturbance from underwater noise. Disturbance of cetaceans can be difficult to detect, but can include: changes in (direction or speed of) swimming or diving behaviour; bunching together or females shielding calves; changes in breathing patterns; changes in vocalisation; aggression, agitation or panic behaviour; certain surface behaviours such as tail slashes and trumpet blows; moving out of an area previously occupied (Marine Scotland, 2020a). The likelihood of marine mammals being impacted by an underwater noise is likely to be affected by the type of device and the number of transducers being used as well as the site-specific sound propagation characteristics and the behavioural, physiological and motivational state of the animals.

Several studies have found that wild harbour porpoises avoid signals from a range of ADD types, both real and simulated, including ADDs used in aquaculture (Johnston, 2002; Olesiuk et al., 2002) as well as ADDs that are not used in aquaculture (Benjamins et al., 2018; Brandt et al., 2013 Mikkelsen et al., 2017). One study found similar avoidance in minke whales to a Lofitech device (McGarry et al., 2017). These studies have invariably considered only short-term displacement, whereas the potential impact of ADD use in aquaculture could be long-term. Short-term displacement does not necessarily lead to long-term displacement. Porpoises were temporarily displaced by Lofitech ADDs used for offshore mitigation, but detections were observed again after a minimum of 133 minutes within 1 km (Thompson et al., 2020). In addition, behavioural response to marine noise may change over time, for example the probability of porpoises being disturbed by pile-driving was found to diminish over time (Graham et al., 2019). Results of a trial using pingers (low amplitude acoustic deterrents) showed a similar reduction in the level of displacement to harbour porpoises (Kyhn et al., 2015). This shows that research seeking to understand population level consequences of anthropogenic marine noise should aim toward understanding the drivers of behavioural response in a range of contexts and attempting to predict the fitness consequences of observed alterations in behaviour.

Different species of marine mammals are known to react quite differently to sound signals (Kastelein et al., 2006). Furthermore, responses also vary between and within individuals and populations (Harris et al., 2017), highlighting the importance of behavioural context in modulating dose–response relationships.

Behavioural responses can be energetically costly, both in terms of additional movement and stress responses, but also in terms of lost foraging opportunities. Harbour porpoises showed increased respiration rates and swim speed (indicative of stress) when exposed to ADD signals in captivity (Kastelein et al., 2015), though this has not been demonstrated outside of captivity. Studies have found similar results for other cetaceans exposed to vessel noise (e.g. humpback whales, Megaptera novaeangliae, Sprogis et al., 2020). Harbour porpoises have very high metabolic rates and therefore energy requirements (Rojano-Donãte et al., 2018) and this may make them especially vulnerable to the effects of displacement if they cannot compensate for increased metabolic costs or lost foraging (Booth, 2020).

Research is required to better understand the level of disturbance and displacement potentially caused by different types of ADDs in different contexts. Specific approaches to address these research requirements are outlined below.

Approach 1: Monitoring the distribution of marine mammal species in a 'real-world' scenario.

Monitoring changes in species distribution around existing finfish farm infrastructure has the advantage of being a true representation of marine mammal interactions with ADDs. Information collected in this way will therefore be ecologically relevant. It has the disadvantage of many confounding factors: sources of noise other than ADDs, potentially complex geo-acoustic environments, potentially low encounter rates of the species of interest and the fact that animals may already have been exposed to noise sources, preventing any assessment of habituation or sensitisation. Controlling the use of ADDs would require collaboration with the site owner and manager.

Approach 2: Use of a targeted approach (not based at a finfish farm), utilising another platform to deploy an ADD.

This could involve controlled exposure experiments in combination with telemetry or visual tracking. Alternatively, this could involve long-term deployment of different ADDs at sea in areas away from fish farms with a network of associated acoustic monitoring devices deployed around the ADD. The advantage of this approach would be that high population density areas could be targeted, increasing the sample size, and subsequent statistical power. A location could be selected that is not deemed to be critical habitat (e.g. not inside an SAC), and potentially without complex bathymetry that may affect sound propagation or animal movement. This approach would also allow for strict control of experimental parameters, allowing for direct comparison between the effects of different device types. The disadvantage of this approach is that results may not be applicable to finfish farm habitats, where the behavioural context could be different, for example with respect to animal motivation and human activity.

Habitat degradation

While several studies have demonstrated harbour porpoise avoidance of ADD signals, none have demonstrated complete habitat exclusion. This may be context dependent, for example if animals are highly motivated to remain in an area of increased underwater noise. If animals are not displaced, there is a risk of negative impacts associated with underwater noise, such as increased stress which has been linked with suppression of the immune system and reproduction, disruption of foraging and social learning, and increased rate of mortality (Atkinson et al., 2015).

Acoustic masking has also been raised as a concern in relation to ADD use (Shapiro et al., 2009). Masking occurs where an introduced sound makes it harder for an animal to hear another signal (the process of one sound increasing the threshold at which another sound can be heard). The transmission characteristics, frequency spectra, duty cycle and number of transducers will be important factors in assessing the potential for masking to occur. The potential for masking will be higher where noise sources are continuous and similar in frequency to ecologically important sounds, which is primarily a problem where anthropogenic noise occupies the same bandwidth as animal vocalisations. The ecological consequences of masking are poorly understood, but thought to be potentially significant for species that rely on sound for navigation, communication and foraging (Erbe et al., 2016). Masking can also affect an animal's ability to hear sounds that they rely upon to avoid danger. Although masking has been raised as a potential concern in relation to ADD use, there have been no published studies of the potential for masking to occur as a result of ADD use.

It is important that the impacts of potential habitat degradation are considered cumulatively, including the potential impacts of other marine stressors and anthropogenic activities such as climate change, fishing, disease, algal blooms, contaminants, and additional threats such as bycatch and ship-strike (Wright et al., 2007), most of which have yet to be quantified.

Understanding the effect of ADD noise on the acoustic environment of marine mammals is complex and would involve prediction and assessment of changes in communication and listening space as a result of auditory masking by the introduced ADD signals. Masking is a complex phenomenon and masking levels are difficult to predict for any combination of source, environment and receiver characteristics (Erbe et al., 2016). Predicting the effects of masking requires an understanding of species-specific audiograms, critical ratios, critical bandwidth and auditory integration times (Erbe et al., 2016). Marine mammals have been documented to have evolved strategies for enhancing the detectability of signals in the presence of masking noise; this has been demonstrated in receivers which can exploit additional acoustic information to reduce the expected effect of masking (Bain & Dahlheim, 1994; Turnbull, 1994) and in senders which can alter the characteristics of their signals in the presence of noise (e.g. Hotchkin & Parks, 2013). Erbe et al. (2016) provides a framework to enable the construction of models of masking to determine the potential limitations of communication in marine mammals in the presence of anthropogenic sound. Pine et al. (2019) demonstrate a method for calculating the effects of masking in terms of the reduction in listening space that occurs as a result of anthropogenic noise emissions. The listening space differs from communication space in that it extends beyond intra-specific communication and also includes the detection of acoustic signatures from conspecifics, prey, predators and/or danger (Pine et al., 2019). It also differs in that prior knowledge of the species-specific auditory filter, gain, detection threshold, signal directivity and duration are not needed; the only species-specific data required is an audiogram. A review is recommended to assess the potential for ADDs to cause masking to the detriment of the most commonly occurring marine mammal species in areas of ADD use. This could include consideration of the methods described in Erbe et al. (2016) and Pine et al. (2019) to quantify the potential for masking to occur to assess its potential significance.

Auditory injury

Marine mammals exposed to intense sound, either instantaneously or over time, have the potential to exhibit reduced hearing sensitivity, termed 'threshold shift. Threshold shifts occur when the hearing sensitivity at a particular frequency (or a range of frequencies) is reduced, either temporarily (temporary threshold shift – TTS), or permanently (permanent threshold shift – PTS). PTS is a form of auditory injury (also known as hearing damage) and is generally considered to be the primary risk from intense underwater sound to marine mammals. TTS is a precursor to PTS, and is often measured in experimental studies as an indicator of PTS risk. The link between TTS and PTS is complex, but well established (NOAA & NMFS, 2018). An offset of 20 dB above measured TTS onset is recommended to predict PTS onset (Southall et al., 2019).

While TTS is not considered as a form of injury in the UK, in the presence of chronic noise such as continuous or intermittent exposure to ADD noise, repeated TTS may also have ecological consequences, for example due to lost foraging opportunities during recovery time.

Captive studies of harbour porpoises and seals have confirmed that a simulated ADD signal can cause TTS under certain conditions. In a study by Schaffeld et al. (2019), a porpoise was exposed to a simulated Lofitech ADD noise at 14 kHz, and TTS onset was found at amplitudes above 141.8 dB re 1 µPa² s Sound Exposure Level (SEL - a measure of cumulative sound exposure), and 155.2 dB re 1 µPa peak-to-peak Sound Pressure Level (a measure of instantaneous exposure). This onset level was significantly lower than the most recent marine mammal noise exposure criteria for continuous noise (Southall et al., 2019). Using the 20 dB offset between TTS onset and PTS onset, this would suggest PTS onset at 162.8 dB re 1 µPa² s SEL (or 158 dB re 1 µPa² s after adjusting for harbour porpoise hearing sensitivity).

In empirical studies with seals and porpoises, TTS is typically observed at frequencies 0.5-1.5 octaves higher than that of the sound stimulus. A harbour seal exposed to a 4.1 kHz pure tone suffered TTS at 5.8 kHz, approximately 0.5 octaves higher (Reichmuth et al., 2019). TTS was observed in a harbour porpoise at 20 and 28 kHz, an octave above the stimulus frequency, and artificial 14 kHz ADD signal (Schaffeld et al., 2019).

The risk of hearing threshold shift to marine mammals from underwater noise depends not only on the source amplitude and frequency, but also critically the length of exposure time (cumulative sound exposure) (Southall et al., 2019). Behavioural responses and avoidance are not well enough understood to reliably estimate the risk of PTS to individual animals, but modelled sound exposure from ADDs indicates there is a credible risk of injury to the hearing of seals and porpoises from cumulative exposure to some ADDs (Götz & Janik, 2013; Lepper et al., 2014). Susceptibility will likely be affected not only by the level of cumulative sound exposure, but also the acoustic properties of the sound stimulus. The risk is mediated by species hearing sensitivity, the frequency of the signal, and for intermittent noise the inter-pulse interval (Kastelein et al., 2014). The cumulative exposure of an animal to a sound signal will be affected by movement patterns and any behavioural response to the sound; understanding movement patterns (especially in areas of continuous or intermittent noise) is therefore highly important (Aarts et al., 2016).

Available information on the likelihood of ADDs causing auditory injury in marine mammals is mostly derived from captive and modelling studies. Research is recommended to translate this information into a real-world approach, to inform commercial use of ADDs in aquaculture.

Approach 4: Individual based modelling

Individual based modelling (IBM) approaches could be used to simulate movement of cetaceans in response to spatially explicit modelled sound, potentially building upon existing IBM work for estimating the movements of harbour porpoises and seals (e.g. Chudzinska et al., 2021; Nabe-Nielsen et al., 2018). Cumulative sound exposure could be estimated for individual animals by making assumptions about movement patterns and behavioural responses to noise. Variables such as ADD source levels, effective duty cycles and propagation characteristics could be estimated based on available data.

There is little information currently available relating to animal responses to ADDs to parameterise such a modelling approach. Suitable data might be obtained from ongoing tagging programs in Greenland and Denmark (Nielsen et al., 2018; van Beest et al., 2018), or the feasibility of investigating a programme of tagging in Scotland could be investigated. Captive animal studies might provide some useful data where it is not available elsewhere, but would not remove the need for long-term field studies. Using such data to simulate a range of possible scenarios would provide insight into the potential risk of hearing damage. This would also allow the parallel prediction of injury risk and (if using a model with an energetic component) the prediction of energetic and life history consequences of disturbance and displacement such as the consequences of re-directed transit or the consequences of exclusion from feeding grounds. This approach could also be used to guide future research priorities by identifying parameters that have the most influence on the severity of impacts. It is likely that any studies carried out under approach 1 and 2 would provide useful data to inform such modelling approaches and parameterise elements of the models.

Population consequences of impacts

Most conservation legislative frameworks require a prediction of the effects of anthropogenic activities on the conservation status of the affected population (for example, EPS licensing), so future ADD use may require further detailed consideration of this, across all of the potential impact pathways. However, for all of the potential impacts on non-target species detailed above, there are significant knowledge gaps relating to our ability to determine population consequences of impacts. This problem is shared with other noise related impacts on marine species, and is not unique to the understanding of ADD impacts. While it may be possible to estimate the number of animals that exposed to ADD sound emissions and experience the different effects (disturbance, auditory injury and masking) using local density estimates, noise propagation models and animal movement models; predicting individual and population consequences of these effects is extremely difficult. The development of approaches to predict the population consequences of noise related impacts is an active area of research and there are developing approaches and frameworks that could be adopted, modified and developed further to improve our understanding of the effects of ADDs on non-target species.

Understanding species' functional use of the area affected by the impact (and the availability of alternative areas for that function) is crucial to our ability to predict the likelihood and consequences of displacement and masking, as well as the likelihood of response movement in relation to the prediction of noise exposure and injury risk. Understanding the nature and degree of observed displacement and its effect on individual survival and reproduction is also important and to translate these into population consequences, knowledge about the conservation status of the population is required.

Only a very limited number of these elements (energetic and fitness consequences of effects, population conservation status, functional use of habitats and availability of alternative habitats) can be quantified or measured at present, which restricts the development of existing frameworks into more meaningful predictive modelling frameworks. However, the development of such a framework in the absence of data is valuable, in that it could be used to determine the most sensitive elements of the process, in order to guide future development of the modelling and prioritise data gathering. This approach could also be used to simulate and test hypothetical outcomes at the population level. For example, simulations to quantify the energetic consequences of different scenarios could be used to inform population-based models (e.g. Harwood et al., 2020). This may provide a hierarchy or ranking of concerns that could identify metrics and triggers for monitoring and adaptive management and would highlight where research effort should be focussed.

5.4 ADD efficacy

Analysis of the available observational data through this project indicated a positive relationship between ADD usage and seal depredation, even when other available factors were accounted for through statistical modelling. As discussed previously, the data do not enable a definitive understanding of the cause of this relationship and alternative approaches are required to draw meaningful conclusions. We discuss three potential approaches to investigate the effectiveness of ADDs in reducing seal depredation. Consideration may also be required of the impact of ADDs on seal presence, in order to assess the effect of seal presence on stress caused to fish.

5.4.1 Additional data collection

Data available for the analysis in this report was mainly at the temporal scale of a month, with some depredation data available at approximately weekly scale. Finer temporal scale data could potentially be gathered on ADD usage and depredation, for example at the daily or weekly level, combined with more explanatory variables related to depredation, which would allow more refined statistical models to be constructed. However, this type of observational data is not designed to infer causation and depending on the strength of any relationships found, it may still be possible to argue that they were caused by some unmeasured factor.

It is sometimes possible to present a convincing causal argument from observational data alone (for example in many medical studies) but a detailed understanding of the mechanism and evidence for rejecting all credible alternatives is required, and the task is more difficult when the effect size is small. Results from the analysis in this report indicate that the overall effect of ADD use on depredation is not overwhelmingly large, but it is important to note that the observational nature of the available data preclude a direct assessment of ADD use versus the same situation without ADD use. To discount the observed positive association between ADD use and high depredation rates, an extensive and detailed dataset would be required. A suitable number of sites would also need to be observed both with and without ADD use. ADDs are either used or not used, which precludes the collection of data from both of these treatments from the same site. This leaves the option of comparing data from different sites, which may not be comparable for underlying reasons such as differing site infrastructure, local seal population size etc. If finfish farm managers are becoming more 'reactive' in their use of ADDs, this may present an opportunity to collect such comparable data, as ADD state (on/off) will be switched more frequently. This factor, combined with the potential for high-resolution automated data collection through online web interfaces, may provide the means for collection of such information in future.

5.4.2 Experimental controlled trials

There are multiple ways that experiments could be undertaken to estimate the effect of particular anti-predator treatments, including ADDs, on seal depredation. Randomized trials, involving controls and replication, are widely accepted as the best way to estimate the effect of an intervention, where it is possible ethically and financially to perform them. An experimental design based on these principles is strongly recommended, and one potential approach is described here.

The aim of such an approach would be to estimate the reduction in depredation frequency that occurs from a single ADD type, deployed in a particular way (the design is readily expanded to include other types, other treatments, etc.). The experimental unit is a stocking period at a finfish farm. The way that farms are chosen to participate will affect the inferences that can be drawn from the experiment. For example, if farms are selected on the basis that they have high expected levels of depredation then inferences that can be drawn from the study about any effect will apply only to farms with high expected levels of depredation. This would not necessarily be a problem if conclusions about ADD efficacy will only be applied to farms with high depredation, and will not be extrapolated to farms where depredation rate is lower. Ideally, farms would be selected at random from the pool of all farms of interest. The most readily available unit of measurement is the number of fish killed by seals, but there may be a more accurate way of measuring the potential impact of ADDs on seal behaviour, such as sighting rate of seals within a certain distance. This latter approach would be required to understand the extent to which ADD use prevents the presence of seals around farms, and therefore eliminates the potential for seal presence to cause stress to the fish.

Multiple finfish farms should be selected to provide replication, and treatments should be allocated randomly to chosen farms (suggested treatment regimes are given below). Those collecting the resulting data would ideally be blind to whether the treatment involves the ADD producing sound or not. In the case of certain ADD types, this might be possible through remote cycling of ADD status via the manufacturers' online portal. Ideally, data should be collected independently to ensure that those collecting the data (depredation rates) are not invested in the outcome.

An example treatment regime would involve treatment periods of one week. This is timed to coincide with a typical inspection routine (mortality removal) and could be adjusted if a different period is more convenient. However, regime timing must remain consistent throughout the study. Each week within a stocking period would be randomly allocated to ADD on or ADD off. During weeks where the ADD is off, all equipment would be left in the water but no sound would be produced by the ADD. Mortality would be recorded each week, with depredation rate quantified by a trained member of staff.

One advantage of this approach would be that additional investigations could be conducted concurrently to gain a better understanding of seal depredation behaviour. For example, depending on the number of sites involved and the timespan required for the different treatments, observers could be stationed on site to track the movement of individual animals. This type of approach has been used successfully in the past by Harris et al. (2014), who used photo-identification of individual seals at coastal bagnets to assess the level of depredation behaviour attributed to small proportion of identified seals (termed 'rogue seals' or 'specialists'). Alternatively, a visual tracking approach was used by Graham et al. (2009) to show patterns of seal movement in response to an ADD, and theodolite tracking was used by Götz and Janik (2015) to record precise movements of seals and non-target species at finfish farms.

To illustrate the strength of an experimental approach we undertook a simple prospective power analysis using a simulation approach. The question was defined as whether ADD use reduced the proportion of months in which depredation occurred. Details of the methodology and full results are given in Appendix section 3.

Results indicate that, based on the specific research question outlined, at least 15 finfish farms might need to be tested with a monthly on/off ADD treatment for the length of one stocking cycle (typically 12 to 18 months) in order to show a significant result. If treatments could be switched at the weekly level instead of monthly, and data could be collected weekly, this would be likely to reduce the monitoring period required considerably.

5.4.3 Adaptive management

A formal experimental setup such as that described in subsection 5.4.2 may be considered impractical for use on a large number of commercial farms. However, it is clear that an experimental approach would be required to resolve the issue of whether, and how, ADDs affect depredation. In this context, the framework of adaptive (resource) management (Walters, 1986; Williams, 2011a; Williams, 2011b) may prove useful. Adaptive management approaches allow exploration of different ways to meet a defined management objective. One or more of these alternatives is implemented and monitored to learn about the effects of management decisions, and then results are used to improve knowledge and adjust management actions appropriately. In the case of ADDs this would allow for ongoing decision-making on permitted use of ADDs in the light of best available science while at the same time designing and implementing regulated experiments as part of a broader management approach and using feedback from results to update the optimal management strategy.

5.4.4 Effect of stress caused by depredation on finfish growth and disease

It is widely considered by finfish producers that the presence of seals causes stress to stocked fish, which leads to reduced growth rates and increased disease (The Scottish Government, 2021). If there is a clear link between seal presence and reduced fish health, even in the absence of direct injury and mortality, then there would be strong case for the use of anti-predator measures which prevent fish from detecting the presence of a predator, as opposed to those which simply prevent the predator from accessing the fish, such as anti-predator netting. No scientific studies have been published on whether or not predatory behaviour by seals has a negative impact on the health or welfare of stocked finfish through increased stress. There is some evidence of predatory behaviour reducing the health of stocked fish based on captive trials with other fish species (Barcellos et al., 2007), but no published evidence is available in the context of predatory seal behaviour.

This question could be addressed either through captive trials or through data collection at fish farms. Captive trials could be split into two parts: quantification of the level of stress created by predatory seal behaviour on finfish, and then assessment of how that level of stress affects finfish health outcomes. Such trials would require ethical assessment. Trials at finfish farms could take advantage of large volumes of 'real-world' data on growth rate, feed rate and / or rates of disease, but linking these with an accurate measure of predation behaviour is more challenging. If reliable metrics could be obtained on the efficacy of ADDs, or any other anti-predator management measure, then this could potentially be used as a proxy for predator presence/absence. For example, mimicking the format of the trial described above, we could compare feed conversion rate during periods of ADD use (assumed to be low rate of stress caused by predatory seal behaviour) against periods of no ADD use. This approach would have the fundamental problem of being observational in nature, and therefore the true cause of any effect could not be conclusively attributed to seal presence/absence, as there may be underlying factors linking ADD use with growth rate.