Types of treatments and rates of usage. The most commonly used chemical was azamethiphos (84.6% of treatments) followed by EMB (70.8%), ivermetin (24.6%), and hydrogen peroxide (16.9%). We can also note an overall decline in usage of ivermectin in the past 2 years. Sequential usage of chemicals is the most implemented approach though all compounds other than hydrogen peroxide were also employed as an only mean of treatment at some sites for a given year. The principle of rotating treatments has proven to be essential in order to maintain efficacies for as long as possible3. The most commonly used combination of products was found to be EMB and azamethiphos with 41.5% of the total number of treatments from 2016 to 2019. To prevent re-infection during the high risk period, when fish are clearing EMB, farmers may apply a chemical bath treatment as soon as juvenile sea lice are detected10. The frequent usage of EMB followed by azamethiphos baths suggests that this is a strategy seemingly effective in NL sites.
The standard EMB treatment regime is 50 μg/kg fish·day-1 for seven consecutive days26 resulting in 0.35 mg/kg for one EMB treatment. EMB usage per biomass calculated in Table 3 has values of 3 and 146 times higher than one EMB recommended treatment when contrasting min and max rates respectively. These amounts are likely related to more than one sea lice parasitic event with fish treated for recurring infestations within one year but they could also be associated with potential resistance patterns. In a review of drugs and pesticides use by the Canadian marine finfish industry in 2016-18 Chang et al.8 noted that the province of British Columbia (BC) had the largest number of farms treated with EMB, while NL used the largest quantity of AI in all three years confirming to some degree our observations on high rates per biomass. These observations cannot replace actual testing of sea lice populations such as in other studies that have elegantly combined both empirical observations and toxicity data on lice5 but they can provide a sign. The effective concentrations of chemotherapeutant needed to remove 50% of lice for resistant strains can be more than 3 to 100 times higher than for sensitive strains depending on chemicals6. Salmon producing areas have documented the emergence of EMB resistant sea lice within 5 to 10 years of the product start of usage5, 6, 27, 28, 29. Within the Canadian context this has been documented in some areas such as in the Bay of Fundy30 with potential drivers for drug resistance in BC10. The North-Atlantic population of salmon lice is genetically very homogeneous31, as larvae can disperse over large distances prior to infecting new fish32. Homogeneity has also been demonstrated for resistance towards the organophosphate azamethiphos with the associated mutation widely dispersed in the North Atlantic population of lice33. However, another study on UK, Norway and Irish sea lice samples suggest that it is possible to discriminate between nearby L. salmonis populations given suitable marker selection approaches, and that such differences have an adaptive basis34. Further work is needed to investigate resistance in the regional context of our study and in Canada in general.
For bath pesticides, Salmosan® is used at a dose of 100 µg/L for 30-60 minutes in well boats and tarps35, 36 while hydrogen peroxide (AI of Paramove®) has a registered dosage of 1.2 - 1.8 g/L for 20 minutes 28,37. The values in Table 3 are expressed per biomass and without having the appropriate details on volume, administration routes (tarp versus well boat) it is difficult to comment on intensity. In addition, we have not found values expressed per weight in the literature to adequately contrast usage.
In term of deposited amounts in the environment hydrogen peroxide had the highest weight of AI (Figure 2). Unlike the other compounds considered in this report, hydrogen peroxide should not at all accumulate in the environment as it splits quickly into water and oxygen, and is therefore considered environmentally benign37. However, recent findings document potential short-term toxicity on non-target species present around farms38, 39, 40. We can note that overall chemical input in the environment was reduced in 2019 (not for hydrogen peroxide) and that this trend has not been consistent within the 4 years. In Norway, Overton et al.3 documented that bathing with pesticides and hydrogen peroxide were reduced from 2015 to 2017, with the introduction of thermal and mechanical approaches to fill the void. In the data available to us, we have no detailed information on mechanical and thermal means though their usage has been described in industry press releases while another non-medicinal alternative, i.e. cleaner fish, considered in this study resulted in lower usage of chemicals. A recent analysis of more than 500 farms in Norway suggests that louse removal efficacy of cleaner fish was patchy in space and time41. The sustainability of cleaner fish usage due to issues in fisheries quotas and welfare have also been of concern29, 42and remain to be better studied in the Canadian context. Additionally, a decrease in the annual production of cleanerfish has been reported in recent years in Scotland43.
Effect of treatments, year, and environmental variables. Site production numbers do not appear as influential predictors in the DTA. This could be influenced by the fact that overall biomass at treated sites were not statistically different between years despite the existence of a range of production values. These values are not reflective of an overall production as there was a slight decline in production tonnage after 2016 44, 45, 46.
The first split in the DTA is determined by some types of treatments but mostly whether sites were treated or not and explains 60% of the variance in the total usage of anti-sea lice chemicals. When examining the sites with lower treatments (n=12 sites) and no treatment at all, we can note that these are not the same locations during the different years (Figure 1A,B,C,D), that they have similar median distances to other sites as well as comparable numbers of fish as for the other 50 treatments. Among these yearly data entries 31.1% of sites had lumpfish for the lower and no treatment group versus only 14.0% in the higher treatment group. This could have influenced the data splitting in the first node but this effect does not fully explain the resulting node as the lumpfish presence/absence is not a predictor in the DTA. Larsen and Vormedal4 noted that some production sites were able to keep lice levels low with fewer treatments, whereas other sites had high numbers of lice despite large numbers of treatments. In our study, only 2 sites have had consistently either no treatments or lower treatments during the 4 years. This highlights the potential effect of other farm-related and/or environmental parameter likely requiring a higher data resolution than what is used in this exploratory exercise.
The year and climate were tested separately and determined similar DTA splits in the data (9.3% of the variability) with effects only on the higher treatment sites (Figure 7,8, and 9). When testing temperature effects on sea lice infestations in wild sea trouts, Vollset et al.24 concluded that “year” had confounding effects on statistical analyses as it likely reflects differences in climate between years as also observed by Shephard et al.25. The year/environmental conditions did not affect the lower and no treatment datapoints with no particular pattern in percentages of treated sites per year.This suggests that the climate effects might have enhanced existing sea lice populations but did not determine the actual presence/absence of sea lice. The higher treatment years 2017 and 2018 were characterized by warmer fall surface temperatures as noted also by DFO (2018)47, a higher freshwater input in spring47,48, and stronger NE wind conditions. Unlike some salinity thresholds49, 50 temperature does not have a limiting effect on louse survival in the ranges observed naturally16. Temperature mediated growth of sea lice51 on the other hand likely affected the severity and recurrence of infestations and consequently the amounts of chemical input. Sea lice have been shown to mostly stay within the surface 10 meters52 especially during the day so the effect of SST ranges are likely significant. Temperature is a determinant of louse development rate and reproduction with seasonal patterns12. Godwin et al.53 also found that the risk of sea-louse outbreaks increases in overall high-sea-surface temperature years pointing to the importance of including climate change patterns. In our study area, the ‘spring-freshet’ discharge due to the melting of snow and ice peaks in April-May and results in water stratification54. Sea water freshening (salinity levels < 27-29‰), has generally a negative effect on sea lice49, 50. Except in HB-BDE, near-surface salinity rarely reaches less than 30‰. In HB-BDE salinities lesser than 30‰ are generally found only at the surface above 5m50. Thus, sea-lice can still have access to favorable salinity conditions below 5m even in years of relatively large river discharge55. Stratification plays a role in sea-lice behavior enhancing their drive and ability to find a host below the halocline as showed by a number of studies56, 5758. Crosbie et al.58 recently reported interesting differences in salinity tolerances between lice stages with strong aggregative behaviour towards the halocline layer58. These differences would allow the two stages (nauplii and copepodids) to survive despite limiting conditions at surface (0-5m) such as in HB-BDE with nauplii potentially occupying depths below 5m until they reach the copepodid stage. A modelling study59 integrating the findings of Crosbie et al.58 show that avoidance of water with low salinity need to be parameterized using probability functions while ensuring not to use too strict salinity avoidance scenarios for the nauplii stages. In term of winds, given the main orientation of the bays within a SW-NE axis, a stronger effect of the wind can be expected on the circulation when winds are blowing from those directions22, 60. Hence, stronger winds from the NE could have resulted in larger water movement contributing to more sea lice transmission amongst sites in 2017 and 2018.
There is no significant effect of median distances to adjacent sites. However, we can note a tendency for higher chemical usage in sites that are closer together (Figure 10). Statistical models of lice epidemiology show that the risk of sea lice infestation increases with the number of farms in an area61,62,63. Median distances as evaluated in this study including the total number of fish on a farm might not have adequately captured the site connectivity for effects to be reflected in the DTA results. The number of fish on individual farms may be less important in determining sea lice numbers than the total number of fish on all farms in an area64. Distance alone might not equate transmission between farms as the magnitude and direction of connections influence outcomes with regions classified as net sea lice exporters while others as net receivers of particles62, 65, 66. NL observations have shown that small scale variability is present in the bays22 potentially having an impact on local lice transmission requiring the need to refine the understanding of connectivity and hydrodynamic processes. In addition, the length of the computed time window even if reflective of the potential for water movement between locations might not be relevant74 to capture adequately infection potentials of sites as Lepeophtheirus salmonis need a few days to grow from the Nauplius stage to its infectious copepodid stage with temperature inversely affecting the length of time67. The planktonic larvae of L. salmonis are non-feeding, and survive on the finite energy reserves of their yolk sac. As the copepodid ages, its energy reserves become depleted68. A free-swimming copepodid may survive for many days but its ability to infect a host will likely be reduced over time74. The energy reserves declined sharply in copepodids 5 and 7 days postmolt when compared to those aged 1 and 2 days68. The one day water mass boundaries are likely inadequate to accurately capture lice dispersion potential.
Data uncertainties. This analysis is based on 4 years of data; despite being informative this time-series does not allow to document long-term trends. Similarly to others2, 3, 4, 5, 6, 8, 16, 53 it is important to stress the importance of having access to comprehensive long-term datasets to better inform on sea lice and chemical usage. In order to learn from past mistakes, the understanding of patterns preceding for example resistance development in salmon lice is crucial5. Some countries such as Norway having an unprecedented transparency in its aquaculture industry with regards to operational details5. It is important however to emphasize the need for transparency in the global industry and the reporting of trends in chemical usage wherever it is feasible. Data should include precise information on all treatments including mechanical approaches, mortality, site environmental conditions, standing biomass and more importantly site lice counts.
We have selected rates of active ingredients per biomass and not the number of treatments as a dependent variable in the exploratory DTAs considering the the lack of precision in defining a treatment. The AAR report frequency of treatments as the number of treatment periods, specified in prescriptions, over which drugs and pesticides are to be used on a farm. For example, a single 7 day prescription represents a frequency of one (DFO 2016)69. A 7 day prescription could be related to an entire site or a cage with no information on the number of fish treated for every prescription on the open data. Similarly to other authors3,5 we have chosen to focus analyses on the expression of data per biomass despite uncertainties as this remains the best approach to understand trends in drug usage. Standing biomass at treatments were calculated using the thermal unit growth initially considered for the evaluation of salmonid growth in hatcheries and used extensively on such species70, 71, 72. Caution is required when the approach is applied to different temperature ranges especially with temperatures far from optimal growing conditions71. Industry specialists manipulate growth patterns to adapt to the final type of product they are marketing73. Without precise data on feeding rates, the growth regimes as used in this study might be underestimated while the absence of exact mortality times may have led to an overestimation of fish numbers. Therefore we have only discussed the ranges (min-max) of AI amounts per kg.
Considering that multi-chemical approaches are dominant, summing multi-compound usage on a site and by biomass allowed a better characterization of chemical input. Sums are used as proxies that can only be partially linked to a number of lice as they are also associated with therapeutic efficiencies and resistance. In the absence of access to site specific datasets and similarly to other authors14, 51, 63 we have used regional SST and proxies for salinity measurements and water currents. These environmental parameters are indicative of the relevant features that could influence sea lice but cannot replace actual salinity and current measurements at sites74 and have a very coarse resolution.