Timing is everything: Survival of Atlantic salmon Salmo salar postsmolts during events of high salmon lice densities

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Journal of Applied Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society †Deceased. 1Institute of Marine Research, Tromsø, Norway; 2Norwegian Institute for Nature Research (NINA), Tromsø, Norway; 3Norwegian Institute for Nature Research (NINA), Trondheim, Norway; 4Department of Biology, NTNU Center of Fisheries and Aquaculture, Trondheim, Norway; 5UiT, The Arctic University of Norway, Tromsø, Norway and 6Institute of Marine Research, Bergen, Norway


| INTRODUC TI ON
To complete its complex life cycle, wild Atlantic salmon Salmo salar postsmolts migrate from the rivers to their feeding grounds in the sea during spring and return as mature adults 1-4 years thereafter to their native river. The survival of the Atlantic salmon during its entire marine migration is typically (much) less than 10%, and most of the mortality occurs shortly after the fish has left the rivers (Thorstad et al., 2012).
Salmon farming has become a major industry in Norway during the recent decades. Atlantic salmon is farmed in net pens in the fjords and along the coast and serves as a host to the parasitic salmon louse Lepeophtheirus salmonis (Krøyer, 1837). This ectoparasite has pathogenic impacts on Atlantic salmon by feeding on their blood and tissue, causing skin lesions, osmoregulatory challenges and physiological stress responses in the host. The pathogenic effect is a well-studied topic (Costello, 2009;Thorstad et al., 2015;Torrissen et al., 2013).
With an estimated stock of 386 million farmed salmon (January 2017;Statistics Norway, 2017), compared to about 0.5 million returning wild Atlantic salmon (Anon, 2018), the number of hosts for the salmon louse has increased by several orders of magnitude since the early eighties (Heuch & Mo, 2001). Moreover, the stated political aim is to increase the production of Atlantic salmon and trout with 500% by 2050 (Vollset et al., 2017), which will scale up already documented negative effects of salmon lice. These problems have led to a new regulatory framework, the socalled 'traffic light system' where green, yellow and red lights represent potential increase, stagnation or decrease in the volume of produced fish, respectively, in 13 predefined production zones along the Norwegian coast. The colour-coded impact categories come from a single indicator, i.e. from the effect of lice on wild salmon within each of the independent production zones: 'Green light' is used when 0%-10% of the wild population of salmon is likely to die because of lice, 'yellow light' is used at mortality rates from 10% to 30% and 'red light' is used at mortality rates >30%. Fish farms are in addition obliged to keep lice levels below 0.2 adult female lice per salmon during the smolt migration period (the rest of the year has a threshold of 0.5 adult female lice).
A mature female salmon louse carries two eggstrings with a total fecundity of about 500-1,000 eggs per brood, for farmed and wild Atlantic salmon respectively (Heuch & Mo, 2001). Hatched eggs develop into infective copepodids that may be transported over long distances with the water currents . Infective lice copepodids have a life span depending on the ambient temperature (Samsing et al., 2016), and at 10 degrees they can drift for 17 days before they need to find a salmonid host, otherwise dying of starvation. Thus, it is likely that they are spread in most of the migration route of wild salmon smolts, and consequently, aquaculture may negatively impact wild Atlantic salmon. Previous studies have identified salmon lice as one of the two largest threats to wild salmon in Norway (Forseth et al., 2017;Taranger et al., 2015).
To study the effects of lice, a series of field experiments with parallel releases of treated versus untreated Atlantic salmon smolts, have been performed in Norway and Ireland (Gargan, Forde, Hazon, Russell, & Todd, 2012;Jackson et al., 2013;Krkošek et al., 2013;Skilbrei et al., 2013). The results are conflicting, but a meta-analysis of all studies from Norwegian systems confirms that lice have a negative effect, but primarily in years when the natural mortality is high. When baseline mortality was high, the positive effect of antiparasitic treatment was high: risk ratio (RR) 1.77-meaning that 1.77 times more fish survive and return in the treated group compared to the control. When baseline mortality was low, no significant difference could be detected (RR ~1.00; Vollset et al., 2016). Thus, conflicting evidence and the lack of a clear link between infestation pressure from lice originating from fish farms and direct effects on e.g. mortality in wild Atlantic salmon makes the scientific controversy still largely unresolved (Vollset, 2019;Vollset, Qviller, Skår, Barlaup, & Dohoo, 2018).
Our study site, the Hardangerfjord system, is a 160-km long fjord on the south-western (SW) coast of Norway. This area is a hotspot for salmonid aquaculture industry. In Hardangerfjord, it has been documented that the proportion of returning fish is lower for fish that originate from rivers further away from the coast, i.e. deeper into the fjords, which may be related to longer time periods of exposure to lice or predators in the environment . Moreover, the timing of the migration of the salmon smolt seems to be crucial. Early migrating fish meet a much lower infestation pressure from lice than late migrating fish (Kristoffersen et al., 2018). This is related to the seasonal dynamics of the louse, which have a temperature-dependent population boom in late spring/early summer (Samsing et al., 2016).
Hardangerfjord was from 2010 to 2017 subjected to synchronized fallowing in order to control lice infestations on both farmed and wild salmonids. Accordingly, all farms in the outer part of the Hardangerfjord were emptied of farmed Atlantic salmon in March 2013 but had full production in 2014 (Guarracino, Qviller, & Lillehaug, 2018;Halttunen et al., 2017). We took advantage of the expected low-to-high lice density shift in the years 2013 versus 2014 and performed a large-scale experiment in the river Etne in Hardangerfjord. We released n = 29,817 Atlantic salmon smolts, using first generation hatchery-reared smolts originating from river Etne brood stock. Our experimental design combined four randomized controlled trials (RCTs) over 2 years (May and June releases nested within 2013 and 2014) with 50% of the smolts in each trial treated with prophylaxis and 50% given sham control treatment.
With this setup, we were able to evaluate the effects of manipulated low and high lice infestation pressure on the survival and growth of recaptured Atlantic salmon upon their return to the river as adults. Moreover, a National research platform with dedicated staff and a fish trap (Resistance Board Weir) with a capture efficacy for wild salmon at about 90% (Skaala et al., 2015) was operational in the Etne River from 2013, (Skaala et al., 2015), minimizing potential capture bias.
We hypothesized that higher lice density causes increased mortality in sea run Atlantic salmon (H 1 ), and higher lice density causes a reduced growth rate in returned Atlantic salmon (H 2 ).

| Study area
The study was carried out in River Etne, draining in the outer parts of Hardangerfjord, in Hordaland county, western Norway ( Figure 1).
The Hardangerfjord is among the most intensively used areas on the Norwegian coast for salmon production, with a standing stock of farmed Atlantic salmon of about 80,000 and 95,000 metric tonnes in 2013 and 2014 respectively (Fiskeridirektoratet, 2019). For further details on the study area see Halttunen et al. (2018).

| Experimental design
The experiment started in 2013 and was replicated in 2014; two groups of Atlantic salmon were released in May and June, each year (Table 1). All fish were released close to the mouth of River Etne.
Returning adult individuals were caught in the trap in River Etne after 1-4 years at sea.
Fish used in this study were first generation, 1-year old hatchery-reared Atlantic salmon postsmolts produced from eggs and sperm stripped from broodstock caught in River Etne. Fish were reared at Matre Research Station (IMR) and made ready for release in salt water.
Prior to release, all salmon smolts were tagged using coded wire tags inserted in their snout, which enable fish identification to (a) treatment/control and (b) timing of release. In addition, all fish had their adipose fin removed to enable us to distinguish experimental fish from wild fish in the trap on return to the river.
For the prophylactic antiparasitic treatment, we used a 30-min bath of Substance EX (Pharmaq), hereafter termed SubEX, at a concentration of 2 p.p.m in oxygenated water. This treatment was applied to 50% of the fish, randomly selected, securing a balanced design. SubEX protects the fish by preventing attached copepodids to develop into the next life stage for up to 16 weeks after treatment (Skilbrei, Espedal, Nilsen, Garcia, & Glover, 2015). Identical (sham) treatment was performed on the control fish. This process was performed 3 days before each of the four releases to allow recovery of the treated fish.
After tagging and treatment, fish were transported in closed oxygenated tanks to Etne by car to a 5 m 3 cage in the sea, close to the outlet of River Etne. The fish were kept in the cage for approximately 48 hr before they were released by lowering the net in the cage.
The release was done by night to reduce predation from birds. Prior to release a sample of 30 fish (randomly picked from the net) were killed to measure length and weight.
From 2014 to 2017, i.e. 1-4 years after release, all experimental fish returning to River Etne were caught in the fish trap and killed (wild Atlantic salmon not belonging to the experiment were released above the trap). Data on body length, weight and sex were registered at the return date.

| Estimation of lice infestation pressure
Salmon lice densities were estimated based on sentinel cages (Bjørn et al., 2011)

F I G U R E 1
Map of the study area. Blue colour shows that the outer management area were farms that were fallowed in March 2013. Red triangle denotes the outlet of River Etne, green area shows the area protected from salmon farming (National Salmon Fjord), red dots denote salmon farming sites and black fish symbols show sentinel cages used in the study TA B L E 1 Summary of released salmon smolts and sample sizes for treatment (prophylaxis) and control groups in the four trials. Fish weights in gram ± SD

| Risk ratio
The RR or relative risk quantifies how much more likely the treated group is to return to the home river, compared to the control group.
We analysed differences in return rates between treated and nontreated fish, for each of the four experimental releases, with the following formulae: where ET is the number of return events (E) in the treatment ( show higher returns of the controls. We calculated confidence intervals for the RR with the formulae: where n 1 and n 2 = sample size of treated and non-treated fish released, respectively; x 1 and x 2 are the sample size of returned fish in the treated and control groups respectively. For 95% CIs we used z = 1.96.

| Survival probability
The survival probability (probability of return) was modelled by logistic regression: where Returned fish represents the probability for surviving 1-4 years in the sea and returning to the river (1 for returning fish, 0 for non-returning fish), Lice Infestation Pressure is the estimated environmental infestation pressure (standardized with mean = 0 and SD = 2) of lice and Treatment is prophylaxis against lice versus control. We also tested whether Releaseweight (average fish weight for the group at release) was a significant covariate in the model. As Releaseweight was a non-significant covariate (Estimate = −0.0054, Z = −1.471, p = 0.14), and did not improve the model (using Akaike Information Criterion), we used a simpler model without this factor. For model validation, we inspected residuals and re-run the model excluding one outlier fish.
However, as the results were practically the same, we decided to include all data points.

| Growth at sea
The growth of the fish during its sea migration was evaluated with a linear regression model: where Weight is individual fish body mass at return, Lice Infestation Pressure is the environmental lice infestation pressure (standardized with mean = 0 and SD = 2), Treatment is prophylaxis or control, Seawinter is the number of years at sea before returning to the river (standardized for 2 SW fish by subtracting 2 from the number of seawinters) and Sex differentiates males from females. Fish that spent four winters at sea were excluded from the analysis since these were only observed in one of the trials. We standardized Lice Infestation Pressure and Seawinter in order to have comparable effect sizes between factors and covariates in the model (Schielzeth, 2010). For model validation, residuals were inspected visually (vs. fitted values and leverage, quantile-quantile plot, scale-location). We also re-run the model without two potential outliers, but decided to include all fish in the dataset.
Statistical analyses were carried out in r statistical package version 3.5.1 (R Developmental Core Team, 2019).

| Salmon lice infestation pressure
Data from the sentinel cages showed that the lice infestation pressure In 2014, when the outer part of the fjord (including the migration route of salmon smolts from River Etne) had full production of salmon, much higher densities of lice were present in this area (Figure 3, lower panels).
Again, the lice density increased from May to June (Figure 3).

| Survival, duration of ocean migration and risk ratio
Both (2) glm(Returned fish ∼ Lice Infestation Pressure

lm(Weight ∼ Lice Infestation Pressure
Lice Infestation Pressure, Treatment and the interaction between these were all significantly contributing to the probability of return of adult salmon (p < 0.0001 for all, Table 2). Increasing lice infestation pressure had a negative effect on the probability of return. Treatment had a positive effect in the centre and at high lice densities, but the interaction effect with Lice indicated that Treatment was beneficial for the fish at high lice densities but negative at low lice infestation pressures.

| Growth at sea
The weight of returning salmon increased approximately linearly with increasing number of winters at sea and the fish added about 2-3 kg of body weight per year ( Figure 6). We could not trace any effect of treatment on the size of returning fish (p = 0.58, linear regression, Table 3).

| Lice-induced mortality
We present a unique documentation of mortality effects on Atlantic salmon, caused by salmon lice. Salmon smolts (unprotected control fish) that were released by their native river and exposed to high lice density (June 2014) suffered a 99.97% mortality rate. This was much higher than smolts in the paired release group protected with proph- risk in migrating Atlantic salmon smolts (Kristoffersen et al., 2018;Vollset, 2019), and also illustrate that the ultimate consequence of a  (Kristoffersen et al., 2018).

| Effects on growth
There was no overall effect of treatment on the weight of returning fish. We thus rejected our hypothesis that salmon lice density caused a reduced growth rate in Atlantic salmon (H 2   . Hence, the fish may still respond with stress reactions to attaching or attached lice, potentially leading to reduced growth rates at high lice densities, i.e. as observed in our data. Further studies are recommended to clarify this issue. The interpretation of results from studies of both mortality and growth in the same groups of fish is not straightforward. Negative effects on growth caused by lice infestation in salmonids may be masked by size-selective mortality (Thorstad et al., 2015). Given our high mortality rates, any size-selective mortality may cause a bias in the growth data. Moreover, it is likely that a stress factor which can cause mortality to an individual also may reduce the growth rate in the same fish. Therefore, negative effects of lice on the growth of the fish are easily masked. Reduced growth rates in Atlantic salmon, due to lice infections, are shown previously, both from field and laboratory studies (Skilbrei & Wennevik, 2006;Skilbrei et al., 2013;Susdorf et al., 2018;Tveiten, Bjørn, Johnsen, Finstad, & McKinley, 2010), but depend on marine survival (Vollset, Barlaup, & Friedland, 2019).

| Challenges in the study design of lice-induced effects on salmonid fish
There

| Toxicity of treatment against lice may have caused biased mortality estimates in previous studies
An interesting observation in our study was that twice as many fish against lice. One laboratory study showed no effect on fish growth after treatment with SubEX , and Gjelland and co-workers speculated that intracoelomic (body cavity) treatment with emamectin benzoate (another prophylactic agent against lice) induced behavioural responses in sea trout (Gjelland et al., 2014). Importantly, if a prophylactic treatment (SubEX or others) is toxic to the experimental fish, research may underestimate the real effects of lice. For example, the meta-analysis performed by Vollset and co-workers, using data from 118 release groups and more than 650,000 individual fish, found no effect of treatment (RR ~1.00) when the baseline survival of the fish was high (Vollset et al., 2016).
We argue that this result may be systematically biased by a potential toxic effect of the treatment. In other words, a real and significant mortality to the Atlantic salmon smolts, caused by a moderate lice density, may not be observed in experiments since the effect is masked by a similar mortality caused by the chemical treatment. This issue merits further investigation.
Vollset and co-workers have shown that lice may cause a delayed return in Atlantic salmon, which also alter the age-distribution in spawning populations (Vollset, Barlaup, Skoglund, Normann, & Skilbrei, 2014). Such effects could be caused by selective mortality in early maturing Atlantic salmon individuals, or, perhaps more likely, that lice infestations lead to reduced growth, which delay both maturation and the return to the river.
Our results (within-year and within-treatment-group comparisons, both years) show that a higher proportion of the fish spent more years at sea when they were released at high lice infestations (June), i.e. giving some support to the hypothesis that high lice infestations can delay the return of sea run Atlantic salmon.

| Size and efficiency of protected areas
Interestingly, the Etne fjord, where we released our experimental smolts, is a protected 'National Salmon Fjord' without any aquaculture production. This is clearly reflected in the low lice infestation pressure observed inside the Etne fjord, compared to other parts of the outer Hardangerfjord (cf. Figures 2 and 3). However, as the Etne fjord only covers a minor part of the migration route of the salmon smolts on their way to the open ocean, the protection status can be of limited value or even totally misleading.
Our data document that experimental smolts, which migrated out during a high lice infestation pressure, had an extremely low probability of survival despite being released in a protected fjord.
This illustrates an important argument on a management level: protected areas need to cover a significant part of the area where the organism under protection experiences relevant stress factors (Bjørn et al., 2011;Serra-Llinares et al., 2014

| Stronger lice-induced effects with longer migration routes?
In River Guddal, situated in the central part of the Hardangerfjord, about 30 km North of River Etne, the smolt migration of Atlantic salmon has its peak in mid-to-late May (Skaala et al., 2019). Our results are also relevant for smolts migrating from the inner rivers in the Hardangerfjord. Wild Atlantic salmon from these rivers will likely use longer time for their migration and thus arrive at the outer region of the fjord later than smolts from rivers closer to the fjord outlet. Consequently, Atlantic salmon populations migrating long distances in the fjord are expected to be more seriously affected by lice. For example, in the Vosso River, lice-induced mortality has been estimated to surpass 30%, which illustrates a more general trend: wild Atlantic salmon populations from the inner part of the fjord have lower population densities .
This supports the hypothesis of increased lice-induced mortality in populations with long fjord migrations.

| Timing is everything
To improve management, it is now crucial to understand the popu-   to expand the production of salmon and trout by 500% within 2050 (Vollset et al., 2017).

| Conclusions and policy recommendations
Nevertheless, the lice challenges can be solved. For policy recommendations, we argue that there is a great potential for limiting the negative effects of lice by spatial separation. Fish in aquaculture can be isolated (totally or partially) from both the lice and from wild salmon fish: (a) salmon aquaculture may use closed farming pens (on land or in the sea), which will give near total control over the lice; (b) salmon aquaculture may use improved constructions of farming pens that limit the exposure of farmed fish to lice, for example by using pens with 'snorkel system' that segregates lice and fish vertically, or by semiclosed pens with protective skirts; (c) farming may also be moved away from vulnerable habitats for wild salmonids (fjords and near coastal areas); e.g. by increasing the size of protected areas, or move pens out at sea were the volume of water will dilute and thus reduce, but not solve, the problems with lice.

ACK N OWLED G EM ENTS
Thanks to Per Tommy Fjeldheim and other workers at the field station at Etne. This study was financed by the Norwegian Research

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available via the Dryad Digital Repository https://doi. org/10.5061/dryad.zw3r2 285d (Bøhn et al., 2020).