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

6 Conclusions

6.1 Extent of ADDs

The number of farms with ADDs installed on site has increased between 2014 to 2016 (with a peak between 2016 and 2018) and reduced slightly from 2018 to 2019, but the number of farms who reported having ADDs switched on has remained relatively consistent. This suggests that more sites may have devices installed, but not actively used.

Since 2016, when the number of farms with ADDs began to plateau, the number of transducers installed on finfish farms continued to increase until the end of 2019, but this rate of increase has slowed from 2018. This could, in part, be due to the uptake of ADD models that generally operate with low numbers of transducers per site. The effective duty-cycle of an ADD system is greatly increased by the use of multiple unsynchronised transducers, and the use of multiple transducers could increase the peak received amplitude of sound around a finfish farm, but this effect is complex.

There was an increase in the prevalence of sites using multiple (different) device types between 2014 and 2017, but the total number of sites remained relatively low, and this trend reduced again after 2017. Conversations with farm managers suggest that multiple ADD types may be used simultaneously to produce an increased deterrent effect, for example the novel sound signal from the second ADD might have a stronger effect, or that one type may replace the other if it is found to be ineffective for a period of time. There is currently no published scientific information to explore the validity of these theories.

There are regional differences in the use of different types of ADD, which could be caused by the operation of different finfish producers. Companies operating exclusively in the Northern Isles report to be more likely to use the Ace Aquatec RT1 device in the most recently available data (2019/2020). Whereas companies operating mainly on the west coast and Western Isles, report to be more likely to use the Airmar/GaelForce or OTAQ devices.

There are some differences between the proportion of time that ADDs are switched on across the different regions, but no consistent patterns or trend. Differences could be caused by the types of device used in different regions or the differences in management strategies of fish farms between regions. Despite changes to the proportion of sites with ADDs, and changes in the proportion of time that ADDs are switched on, there has been a relatively consistent number of farms actively using an ADD since 2014 (in the region of 80 and 100 farms). Alternative management measures such as anti-predator nets could also be having an effect on the number of ADDs in use.

The decrease in the proportion of transducers which are actively being used suggests there may be an increase in the active, or responsive, use of ADDs; only using them when depredation is ongoing or when the threat is considered to be high. This may indicate the start of a more active management approach to ADD use, potentially caused by increased awareness of farm managers to the potential impact on sensitive non-target species. If so, this would have the effect of reducing the overall amount of noise pollution created. There is currently little understanding of how the likelihood of disturbance to non-target species is mediated by different usage strategies of ADDs, such as duty-cycling and triggering.

While the available data have allowed general conclusions on usage trends in time and space, they do not explain why these changes have occurred. There is very little fine-scale information on temporal patterns and strategies of ADD use which would be required to fully understand the nature and extent of noise emissions from ADDs and to be able to predict potential effects.

The availability of data from automated systems developed by ADD manufacturers may be a useful way of collecting and providing detailed records of patterns of ADD use to inform future assessments.

The observed changes in the use of ADDs over the time-period covered in this report may have impacts on sensitive non-target species in Scotland. The large number of transducers installed on some sites is likely to increase the effective duty cycle and potentially the peak received amplitude. The increased prevalence of lower frequency devices replacing higher frequency ADDs may reduce the potential for impacts on high-frequency sensitive species (dolphins and harbour porpoises), but evidence is not available to quantify relative sound exposure levels between different devices. Additionally, this change may lead to increased impacts on low frequency species, such as minke whales. This depends crucially on animals' behavioural response to noise, which is poorly understood at present.

6.2 Analysis of efficacy

From data analysed in this project, seal depredation of finfish farms appears to be a common occurrence, reported to occur in 39% of the months that farms were stocked, and in 71% of stocking periods. The median number of fish reported as being killed by seals per finfish farm in months with depredation was 111 fish; with 5% of these farms having reported depredation of more than 1,000 fish.

Use of ADDs in the dataset was also common, being recorded in 53% of observed months, and 61% of stocking periods. Where they were used, they were typically recorded as being kept on for most of the time, for example in 70% of stocking periods where ADDs were used, they were reported to be turned on in >90% of the months within that stocking period.

In our exploratory analysis, we found a consistently higher frequency and level of depredation associated with ADD use. For example, depredation occurred in 47% of months where ADDs were used and 23% where they were not. Mean depredation was 3,177 fish over the entire stocking period when ADDs were used for >90% of the time and 2,261 when they were not used. Attempts were made to account for geographic, temporal and farm-level factors through a set of statistical models, looking at the data at three different levels of temporal aggregation: month, stocking period and depredation event. Broadly, we found that ADD usage was still associated with increased frequency and level of depredation at all three temporal levels) even after accounting for the other available variables. The only exception was that no relationship was found between ADD usage and the total mortality per depredation event once the length of the event was accounted for, although ADD usage was associated with longer depredation events and these in turn were associated with greater mortality.

The observed positive relationship between ADD use and predation could possibly be attributable to ADD usage causing more seal depredation events, which would align with the so-called 'dinner bell' theory. However, it should be considered that ADD usage may simply be associated with other factors that cause seal depredation. A likely explanation for these findings is that some other factor(s) are linking ADD usage with high depredation. For example, that ADDs are used responsively when depredation is first detected, or that they are used pre-emptively when depredation is anticipated based on previous experience or local knowledge. In either case, ADD use cannot be completely effective, but the level by which they reduce predation, compared to what it would have been without ADD use, is unknown. We did not find strong evidence that ADDs were being used in response to seal depredation: for those stocking periods where ADD use was less than 90%, ADDs tended to be used less before the first recorded depredation (39%) than after it (59%), but the same pattern of increasing use of ADDs over time occurred in stocking periods where no depredation was recorded (e.g. 41% usage in months 1-3 and 58% thereafter).

One option we explored is whether sites not permitted to use ADDs could be viewed as a quasi-experiment (i.e. an experiment without randomisation). In such a situation, the level of depredation in sites, stocking periods or months where ADD usage was prevented could be taken as representative of the baseline level. This could then be compared with depredation levels where ADDs were allowed (whether used or not). The frequency of depredation was generally higher (47% of months) when ADD use was restricted compared to when it was not (37%), and this was supported by the modelling, which showed a positive association (after accounting for other known variables) between ADD restrictions and the monthly level of depredation, the number of depredation events and level of depredation within a stocking period, and the length of the depredation events. This would tend to suggest that restricting ADD use may lead to higher rates of depredation. However, the assumption that locations and times where ADDs are restricted are representative of background levels of depredation may not be valid. For example, ADDs may be more likely to be restricted near to conservation areas and these areas may also have higher seal density and therefore higher expected rates of depredation.

There were differences in depredation frequency and level associated with the different types of ADD. Wherever enough data were present, ADDs models were significantly positively associated with depredation in one or more of the models – for example in the monthly model of depredation frequency, Airmar, Ace Aquatec US3, Terecos, OTAQ and MohnAqua all had statistically significant positive coefficients.

The major factor limiting inferences that can be made from these data are that they are observational rather than experimental. However, other limitations of the data analysis should be noted.

In our analyses the response variables we used related to the frequency of depredation (typically proportion of months) and level of depredation (number of fish killed). However, fish early in a stocking period are typically less valuable than those later on, and so a useful third metric to include in future analyses would be the economic value of the losses.

The relatively coarse temporal resolution means that fine scale-details may have been missed. Stocking periods were delimited by having a clear month with no stocking. Depending on the length of the fallow period between stocking periods, two separate stocking periods could have been recorded as a single longer one. A similar issue exists for depredation events, although it is more ambiguous what the correct definition of a depredation event is since, unlike stocking periods, each incidence of depredation is a discrete event, and so there is no natural clustering. Additionally, the level of depredation that could be considered problematic is not a fixed value (e.g. a particular number of fish) but will vary depending on the size of the fish and other factors. Nevertheless, it was observed that even at the month level there were periods where no depredation occurred interspersed with months of continually recorded depredation. In the statistical modelling, some of the final models did not fit the data well, either explaining a low percentage of the overall variation in the data, or with response variable distributions that produced a residual distribution different from that expected. Overall, it is unlikely that any of these issues affected the overall inferences drawn from the models.

Fundamentally, it is difficult to attribute causation to the observed pattern based on purely observational data unless the magnitude of the effect is large. Without adequate experimental control there is always a risk that some hidden variable that is not measured is causing the association. We were not able to differentiate between any effect of ADDs in reducing depredation and underlying factors that link ADD usage with increased depredation. Randomised, replicated and blinded experiments are the gold standard, while adaptive management can also potentially identify causation. Recommendations for both approaches are provided in section 5.3.

Data collected in this study were not sufficient to provide conclusions about the impact of different netting materials on seal depredation. New netting materials have been increasingly used in Scottish aquaculture and there is anecdotal evidence that some types are resistant to seal depredation, but no information is available for quantitative assessment. Conversations with site managers suggested that initial effects in preventing predation lessened over time, possibly due to changes in the qualities of the net material, or possibly due to seals successfully adapting behaviours.

Anti-predator nets are still in use in Scotland, although there are relatively small numbers of sites using them. One primary concern cited by farm managers in relation to their use was entanglement leading to the drowning of predators, particularly diving birds and seals (Northridge et al., 2010). Their apparently widespread use in industries overseas suggests that problems with entanglement of animals may have been overcome. Alternatively, it may be that conditions in those industries overseas are somehow different to the Scottish conditions in a way which allows their effective use (e.g. lower rates of tidal flow). Research recommendations relating to netting materials are considered in detail as part of an associated report (Thompson et al., 2021).

6.3 Summary of research priorities

To monitor, manage and measure the efficacy of ADD use and any resulting effects on non-target species there is a need for comprehensive and systematic collection of standardised data on the nature and extent of ADD use in Scottish aquaculture. This would include the following information for each site on at least daily basis: ADD model(s), sound source levels, number of transducers, duty cycle, times of operation. Associated information on depredation rates at the same temporal resolution would also be valuable and these should be collated into a centralised data collection system.

The feasibility of the use of anti-predator net systems which have proven to be successful overseas should be explored, including the consideration of controlled trials at Scottish sites, and research recommendations have been outlined as part of an associated report (Thompson et al., 2021).

The following research priorities have been identified:

Effects of ADDs on non-target species:

• Although several studies have demonstrated that ADDs can elicit behavioural responses in cetaceans, there is limited evidence for broadscale displacement around sites where ADDs are used.
• Research into the potential for ADDs to cause significant habitat degradation, using quantitative modelling approaches to explore the potential for masking to occur.
• The development and application of movement models to predict the risk of auditory injury under a range of scenarios with the potential extension of being able to simulate the energetic consequences of predicted responsive behaviour.
• All of these potential effects require to be placed in the context of individual life history consequences, and ultimately population consequences.

Efficacy of ADDs.

• Site based control-treatment trials would be required for this type of study although there may be difficulties in achieving the idealised design in a way that is compatible with an ongoing industry.
• Adaptive management may be a more practical approach to work with the industry to ensure that any ADD use is managed and monitored in a way that maximises the information available to evaluate effectiveness.

These will require dedicated experimental studies designed explicitly to answer the question of effectiveness of ADDs in reducing depredation:

In addition, there is a requirement for research into the effects of stress caused by predation on finfish health, growth and disease. Understanding the effects of predator presence of fish health will help in the design of control measures to ensure that all negative effects of seal presence and depredation can be reduced though active management.