Behaviour and survival of wild Atlantic salmon Salmo salar captured and released while surveillance angling for escaped farmed salmon

: In many Norwegian rivers, spawning stocks are surveyed for escaped farmed salmon with surveillance fishing by rod and reel after the recreational angling season. However, the benefits of surveillance fishing depend on the ability of wild salmon to return to the spawning stock. To evaluate the impacts of surveillance fishing, we captured, radio-tagged and released wild Atlantic salmon Salmo salar in the River Lakselva, Norway, in a surveillance fishery occurring just prior to the spawning period. Among 39 salmon captured, 36 wild fish were tagged and released, whereas 3 were not released (1 bleeding from the gills, 1 farmed, 1 farmed and bleeding). Survival of fish captured by surveillance fishing was high (95% total survival, 100% catch-and-release sur-vival). Tagged fish were tracked on average 1.2 ± 2.8 (SD) km from the release site at the end of the experiment during the spawning season, not significantly different from the distance moved by salmon radio tagged throughout the summer during a similar interval (15 September to 24 October 2014). Total movement within 3 d of release was inferred to average 1.9 ± 2.1 km, exclud-ing 1 individual that exited the river. Tracking data revealed an immediate behavioural reaction of salmon to surveillance catch-and-release angling, the long-term consequences of which are uncertain. Surveillance fishing may be problematic in rivers with small and vulnerable wild stocks in which a high proportion of the spawning populations is sampled. Surveillance fishing completed with ample time before spawning would be a precautious approach to minimize potential effects during spawning.


INTRODUCTION
Atlantic salmon Salmo salar populations have declined throughout their range and many have been extirpated (Parrish et al. 1998). Yet, Atlantic salmon is among the most abundant fishes on earth because its high economic value has resulted in intense cultivation and the global proliferation of salmon aquaculture (Gross 1998). These farmed Atlantic salmon have been subjected to artificial selection that has in creased the growth rate, fat content and age at maturity, and has reduced stamina, egg size and genetic diversity (Gross 1998, Ferguson et al. 2007). Farmed salmon have been subject to commercial breeding programmes since the early 1970s, and as a result, now display a wide range of genetic differences to wild conspecifics (Ferguson et al. 2007, Karlsson et al. 2011, Glover et al. 2017. Neverthe-less, farmed sal mon may still interbreed with wild salmon, potentially causing significant outbreeding effects (Mc Ginnity et al. 1997, Fleming et al. 2000, Skaala et al. 2012. Salmon that escape aquaculture facilities enter the ocean and may aggregate with conspecifics at marine feeding areas prior to maturing and migra ting into fresh water (Hansen & Jacobsen 2003). The proportion of escaped farmed salmon in samples of Atlantic salmon from Norwegian rivers varied on average between 9 and 18% close to the spawning period during 2006 to 2015 (SBSM 2016). The occurrence of escaped farmed sal mon in Norwegian rivers across many years has resulted in a significant genetic introgression from farmed salmon in many wild populations (Glover et al. 2012, 2013, Skaala et al. 2012, Karlsson et al. 2016). In the most recent of these studies, Karlsson et al. (2016) reported that significant genetic introgression was observed in 77 out of 147 Norwegian rivers analysed. Furthermore, outside Norway, introgression of escapees has also been observed (Crozier 1993, Clifford et al. 1998. A recent study has shown that gene flow from escaped farmed salmon has altered age and size at maturation in wild Atlantic salmon in many Norwegian rivers (Bolstad et al. 2017).
Intrusion of non-native salmon threatens the genetic integrity and viability of wild salmon (Fleming et al. 2000, McGinnity et al. 2003, Skaala et al. 2012. Anglers may be able to recognize, and subsequently remove, escaped farmed salmon by identifying morphological differences, but genetic analyses or scale reading is necessary for accurate determination in many instances (Lund & Hansen 1991, Fiske et al. 2005. Scale sampling programmes from recreational fisheries can provide evidence about the ex tent of farmed salmon intrusion in wild salmon rivers. However, farmed Atlantic salmon are known to enter rivers later in the season than wild fish (e.g. Moe et al. 2016, Svenning et al. 2017, and often after the recreational fishery has closed. Therefore, scale samples obtained during the summer underestimate the extent of farmed salmon in the population. To address this, many rivers have specialized surveys of the spawning stock for escaped farmed fish using rod and reel in the autumn, a time when the highest number of fish have entered the river, to collect scales from a more representative sample of the spawning population in the river. Inevitably, many wild Atlantic salmon are captured by these surveillance fisheries, which are intended to be released unharmed so that they can return to pre-spawning and spawning activities. Survival of Atlantic salmon released by rod and reel is consistently high (Lennox et al. 2017), but there is evidence that recreational angling can alter the behaviour of wild Atlantic salmon during their upriver migration (Mäkinen et al. 2000, Thorstad et al. 2003, Havn et al. 2015. Many Atlantic salmon captured in summer fisheries are captured during the upriver movement phase of migration, which begins after river entry. Fish captured in surveillance fisheries are likely to have completed upriver movement and be in the holding phase of the migration, at or near the spawning destination (Økland et al. 2001). Capture, handling (including scale removal for analysis) and release of wild Atlantic salmon in autumn surveillance fisheries may result in mortality of released fish or elicit similar behavioural responses such as long movements away from the holding sites. Success of surveillance fisheries is contingent on balancing the benefits of enumerating the intrusion of farmed salmon against the potential impacts that could be imparted on wild salmon caught and released close to the time of spawning, which remains poorly understood. To determine whether wild Atlantic salmon captured by autumn surveillance fishing with rod and reel are negatively impacted by catch-and-release angling, we tagged salmon with radio transmitters after capture and scale sampling and monitored in-river movements until the spawning season.

Study site
The study was performed in the River Lakselva, Finnmark, Norway. Lakselva discharges into the Porsanger Fjord and has 45 km available to wild Atlantic salmon. Most of the salmon spawn in river reaches below 2 lakes (Lennox et al. 2016). The recreational fishing season for Atlantic salmon concludes on 31 August, and fishing for sea trout Salmo trutta concludes on 15 September. Surveillance fishing is then conducted until approximately 2 wk prior to peak spawning. For this study, surveillance fishing was conducted between 19 September and 2 October 2016. During this period, water temperatures were (mean ± SD) 9.6 ± 0.6°C (range = 8.4−10.7°C) as measured by a water temperature logger (HOBO Pendant Temperature/ Light Data Logger 64K-UA-002-64) deployed in a shaded area of the river at 9 river kilometres (rkm, i.e. km from the river mouth) at a depth of 2 m from the surface.

Experiments
All tagged salmon were captured by 3 anglers. Sex was visually determined by secondary sexual traits such as development of kype and colouration. Our research was conducted on 19 female (mean ± SD = 98 ± 20 cm total length [TL]; range = 58−116 cm TL) and 20 male (77 ± 21 cm TL; range = 55−113 cm) At lan tic salmon captured by surveillance fishing. Five individuals (65 ± 8 cm TL; 4 male, 1 female) were considered to be freshly entered into the river based on their silver colouration; 1 of these fish had sea lice attached with egg strings, which indicates that it had very recently entered fresh water. Salmon were captured by artificial flies (N = 19) and lures (i.e. metal spoons; N = 20). After fighting for an average (± SD) of 321 ± 270 s, salmon were landed in a knotless mesh landing net (N = 37) or tailed (N = 2) and were rapidly transferred into a water-filled PVC tube by placing the fish into a plastic sling designed to hold fish with enough water to respire. The fish's eyes were covered by a damp towel in the tube to keep it calm. Each fish was measured (TL in cm), sexed, assessed for colour (silver, intermediate or brown) and tagged. Tags were rectangular (21 × 52 × 11 mm, mass in air = 15 g) coded radio transmitters (Advanced Telemetry Systems) in the frequency range 142.144− 142.484 MHz. The tags were attached by passing steel wire (0.8 mm) through hypodermic needles inserted be tween the pterygiophore bones. A white plastic backplate was placed on the opposite side from the tag (see Lennox et al. 2015Lennox et al. , 2016. After tagging, 5 to 8 scales were removed from the fish's right side, posterior to the dorsal fin near the midline using needle-nose pliers; these scales were used to determine the origin of the fish as wild or escaped farmed. Total exposure to air from landing the fish to release was (mean ± SD) 2 ± 2 s (range = 0−7 s). Fish were held in water for 62 ± 82 s (range = 0−6 min) before they swam away volitionally.
The position of each fish in the river was determined at 1 h, 1 d, 2 d and 3 d after capture as well as once weekly until 24 October. Positions were determined using a vehicle-mounted ATS R4500CD coded receiver datalogger attached to either a dipole an tenna (Magnetic Roof-Mount Dipole, Laird Technologies) or a 4-element Yagi antenna for precise determination. To ensure adequate coverage of the river and to note any fish that moved out of the tracking area, directional stationary logging stations mounted with two 6-element Yagi antennas each were also established near the confluence of the river with the fjord at 4 rkm and at 20 rkm, beyond which access by road is limited. Fine-scale positioning permitted the identification of upstream movement of fish, indicating survival and also to calculate movement from the capture site after sampling. Observation by snorkeling or sudden position movements provoked by snorkeling were used to confirm survival (on 6 and 23 October 2016) for fish that were not recorded to move upstream between tracking surveys.

Data analysis
Catch-and-release mortality was calculated as the percentage of fish determined to have died based on the radio tracking after release. Fish that moved upriver immediately after release required an additional upriver movement occurring more than 2 d after release to be considered a survivor. Snorkel surveys were conducted to verify survival of any salmon for which there was doubt about survival. Total mortality was calculated by adding delayed mortality and immediate mortality (i.e. the sample fish that were killed due to extensive bleeding).
Tracking data were imported into ArcGIS software (ESRI 2011) to calculate the distance from the river mouth of each position in rkm. Positions for each fish were then used to determine the extent of downriver movement immediately after release and the final position in the river relative to the release location. Because fish were tracked at fixed intervals in the first 3 d, we could infer the minimum movement in this period by summing the absolute values of the changes in position of fish in the river from each tracking point. Final positions of fish compared to the release site were calculated and compared to the change in position of fish tagged in the summer recreational fishery in 2014 (Lennox et al. 2016). The 2014 data comprised 21 fish that were tracked on 24 October and 15 September, which is a comparable period to that in 2016. However, fish in 2014 were tagged during the summer (13 July to 29 August) and we therefore interpreted movement of those fish as expected movement of fish during this period. We excluded 1 fish from 2016 that exited the river and compared the relative movement of fish in the September− October period for salmon tagged in the 2 experiments with a Welch 2-sample t-test implemented with the t.test function in R (R Core Team 2017).
To determine which factors influenced movement of salmon from the release site within the 3 d fixed interval tracking period, generalized least squares regression was implemented with the gls function in the R package nlme (Pinheiro et al. 2016). The model is based on 32 salmon given that 4 individuals had incomplete tracking records within 3 d. The full model is presented as the final model. Generalized least squares regression was used instead of multiple linear regression because it can incorporate variance structures that account for heteroscedasticity in model residuals (Pinheiro et al. 2016). The dependent variable, in ferred distance moved within 3 d, was log transformed to account for non-normality of residuals and a varIdent variance structure was incorporated in the model to account for heteroscedasticity of the bleeding predictor variable, which was coded as a factor (Pinheiro et al. 2016). In all statistical analyses, significance was assessed at α = 0.05.

RESULTS
Among the 39 Atlantic salmon captured, 2 individuals incurred injury to the gills and were not released (1 of farmed origin, 1 wild). One additional individual was determined to be of farmed origin based on external traits and was also not released (farmed origin of this individual was later confirmed by expert scale analysis); all other individuals were confirmed by scale analysis to be of wild origin. Therefore, the sample of tagged fish consisted of 36 wild Atlantic salmon (mean 87 ± 20 cm TL; range = 55−116 cm). Every salmon that was released was confirmed to have survived the surveillance fishing based on in-river movements made after release, observation by snorkeling and/or by sudden changes in position provoked by snorkeling. Catch-and-release survival was therefore 100%. Total survival of the fish captured during the monitoring was 95%, considering that 2 individuals were killed due to injuries. One individual (1 sea winter 62 cm male, brown coloured at release) was recorded passing by the stationary logger near the river mouth 13 rkm downriver from the release location before it disappeared from the river, and most likely did not spawn in Lakselva. Movements by this fish were determined to be volitional, and not the passive drifting of a fish carcass, based on the periodic up-and down-river movements recorded by the stationary logger.
Three of the fish that were released were bleeding slightly; of these, 2 were captured by lures and 1 by a fly. One of the bleeding fish shook the hook out while in the net, indicating that it was not likely lodged in a criti-cal location, 1 was hooked in both the upper and lower jaws simultaneously, and 1 was hooked in the corner of the jaw.
Within 1 h of release, salmon were on average (± SD) 0.04 ± 0.12 rkm downriver from the release site (range = −0.52 to 0.17 rkm; Table 1), although 17 (47%) were approximately in the same position (± 20 m) as where they were released and 75% were within ±100 m (Fig. 1). One day after release, 33% were within ±100 m of the release location; 1 moved upriver 2.5 rkm and another moved down 2.5 rkm 314  . 2). After Day 2, only 15% were within ±100 m and after Day 3, 19% were within ±100 m. The final position of salmon, taken at the end of October, was on average (± SD) above the release location by 1.23 ± 2.80 rkm (range = −2.98 to 11.11 rkm), and only 1 individual (3%) was within 100 m of the release site (Fig. 3). More fish moved upriver than downriver from the release site, suggesting that movements were volitional rather than the result of being swept downcurrent. The net change in position was not significantly different for the 2016 fish captured by surveillance fishing compared to fish captured in the summer recreational fishery in 2014 (t = −0.89, df = 36.84, p = 0.38; Fig. 4). The farthest point recorded from the release site was on average (± SD) 2.19 ± 2.87 rkm, although the median was 0.96 rkm. The inferred distance moved within 3 d of release was on average (± SD) 1.93 ± 2.11 km. Silver coloured and male fish moved more than brown and female fish, but the differences were only weakly significant (Table 2). No other predictors were significant in the model (Table 2).

DISCUSSION
Physical and physiological impacts of capture and handling result in some mortality of fish captured in fisheries (Arlinghaus et al. 2007, Lennox et al. 2017. It is therefore expected that the surveillance fishery would result in the mortality of some wild fish, and, similar to other studies of Atlantic salmon, that this mortality would be infrequent. Immediate mortality of salmon in the surveillance fishery, resulting from damage caused by the hook, was similar to that in other studies (3−10%; Havn et al. 2015, Lennox et al. 2015. Delayed mortality of salmon in the surveillance fishery was nil, which is consistent with other assessments using telemetry that have generally identified infrequent post-release mortality, with high water temperature being the most important predictor of post-release mortality (Lennox et al. 2017). Water temperatures in September and October are cooler than during the summer and therefore high water temperature is not likely to be a risk factor for surveillance fishing mortality. This does not preclude the occasional immediate mortality event such as we observed here, but in general, mor- Capture and handling are known to cause biochemical disturbances in fish (see Thorstad et al. 2003 for data on Atlantic salmon blood analyses in response to angling), which can manifest as behavioural anomalies or impairment (Mäkinen et al. 2000, Thorstad et al. 2003, Havn et al. 2015. Even though all released salmon survived the surveillance fishery, a majority of them were beyond 100 m from where they were holding in preparation of spawning. This is suggestive of an acute disturbance caused by capture and handling, which is similar to the behavioural responses of salmon to catch-and-release angling in the summer re creational fisheries. Radio tracking of salmon released from an ex perimentally extended re creational fishery in the Dee River also identified high survival of fish captured later in the season and with similar up-and downriver movement away from the release site (Dee River Trust 2010). However, salmon probably do not normally move away from holding sites in this season; radio-tagged salmon caught in bag nets in the fjord and released in July moved 0.36 rkm (males) or 0.80 rkm (females) be tween 11 October and 1 November in the River Alta, which is close to Lakselva (Økland et al. 1995). Al though these surveys were performed later in Alta than our study in Lakselva, the total distances were inferred from more frequent tracking surveys, ob serving less movement than fish captured in the Lakselva surveillance fishery. Salmon that are holding position prior to spawning are likely maintaining a position near their eventual spawning site and not necessarily on it (Økland et al. 2001(Økland et al. , Richard et al. 2014). Movement away from holding pools therefore does not necessarily suggest a loss of territory, especially given that individuals move among pools during the spawning season (Taggart et al. 2001), and a comparison to the movement exhibited by salmon in 2014 indicated no major differences in displacement.
It is notable that 5 salmon captured in the surveillance fishery were silver in colour, suggesting that they had recently entered the river. Most salmon have premature migration, entering rivers weeks or months before spawning to stage near spawning grounds (Økland et al. 2001(Økland et al. , Quinn et al. 2015. Silver salmon moved more than brown salmon, which makes sense given that these fish would be more likely to be in the active migration phase, on the way to spawning grounds, than holding near spawning territories (Økland et al. 2001). However, the movements made by the silver salmon were not unidirectional towards upriver territory and included downstream running. Downstream running and erratic movement are believed to be symptoms of a stress response by salmon (Mäkinen et al. 2000, Havn et al. 2015. Although Thorstad et al. (2007)   catch and release of silver and brown salmon, salmon might have differences in the stress responsiveness at different stages of reproductive maturity. For example, Raby et al. (2013) described an attenuated stress response of Pacific salmon Oncorhynchus gorbuscha and O. keta on spawning grounds, suggesting a change in physiological pathways later in migra tion as fish prepare for spawning. Escaped farmed Atlantic salmon represent an immediate and growing threat to the conservation of wild salmon populations (Karlsson et al. 2016, Forseth et al. 2017, Glover et al. 2017. Scale sampling showed that 5% of the salmon captured during surveillance fishing were of farmed origin, which is concerning given that there are no farms in the immediate vicinity. The genetic integrity of the Lakselva population was recently characterized as being of moderate quality due to genetic introgression of farmed salmon in the wild population (Forseth & Thorstad 2016, Karlsson et al. 2016. Maintaining records of farmed salmon in rivers can assist with longterm tracking of trends in wild salmon conservation, particularly in prioritizing conservation efforts towards rivers where intrusion by farmed salmon is most frequent. Given that the goal of any monitoring activity should be to minimize impacts on wild fish, other methods could be compared to sampling with rod and reel. However, surveillance fishing appears to be a relatively low-impact method of surveying the spawning stock for farmed fish. Although we observed 100% survival of fish that were released, there was some immediate mortality of fish captured in the surveillance fishery, and we observed acute disturbances to the behaviour, specifically movement away from the release location that is probably beyond the normal activity that salmon engage in during this period. In addition, 1 fish exited the river and was probably lost from the spawning population. Ultimately, the net change in position from September to spawning at the end of October was not different from expected movement based on fish tagged in the early season in 2014, suggesting that most salmon released by surveillance fisheries can return to the spawning population. Future research should make a more direct comparison with a control group to determine what the expected movement of fish is during the autumn and establish to what extent they move within the river. Research should also shift to a more direct quantification of fitness by assessing the spawning success of fish released from the surveillance fishery. Chronic stressors have repeatedly been linked to reproductive suppression or failure in sal monids (Weiner et al. 1986, Pickering et al. 1987, Campbell et al. 1992. Recreational angling, however, is an acute stressor, and some evidence suggests that fish exposed to acute stressors, even proximate to spawning, can restore homeostasis relatively quickly, which would attenuate any longer-term effects. Pickering et al. (1987) measured decreasing androgenic hormones in the blood of brown trout Salmo trutta, which recovered within 24 h in captivity. Correspondingly, Davidson et al. (1994) and Booth et al. (1995) calculated high hatching success of eggs laid and fertilized by Atlantic salmon subjected to simulated late-season angling. Although this is the best available evidence as to whether salmon captured in surveillance fisheries would likely contribute viable progeny, impacts of long-term survival are unknown and deserve further consideration given that hatching success is not an ultimate measure of offspring fitness (e.g. Berntsen & Bech 2016).
In the absence of direct measurements of fitness im pairments caused by surveillance fishing, an appropriate precaution would be to maintain a buffer period between the end of surveillance fishing and the initiation of spawning to provide salmon with time to recover. Future studies incorporating a control group will assist with defining an appropriate buffer period, including physiological (e.g. Raby et al. 2015) and behavioural assessments (e.g. Whitney et al. 2016) of the recovery time of salmon after exercise. In general, longer buffer periods will be best but this must be considered against the run timing of farmed fish to ensure a representative sample (Moe et al. 2016, Svenning et al. 2017. Farmed salmon also tend to have a different distribution within the river than wild fish, and have been observed migrating farther upriver than wild fish (Moe et al. 2016); therefore, efforts to refine surveillance fishing methods may be necessary to further ensure representative sampling.
Despite the behavioural disturbance observed among the salmon in Lakselva, capture and release of a small population sample for surveillance fishing is unlikely to affect the spawning in rivers with large stocks. Indeed, any negative impacts may be offset by benefits provided by collecting important information about escaped farmed salmon intrusion that al lows for remediation or restoration and contributes to good fisheries management and long-term sustainability. Where rivers are considered to have vulnerable stocks (e.g. small spawning biomass), surveillance fishing could be more damaging to the population, because it is necessary to sample a relatively high proportion of the spawning populations in such rivers to obtain a statistically justifiable sample size compared to rivers with large wild stocks. Caution should be exercised when implementing surveillance fisheries with river-specific conservation needs and objectives of foremost consideration, particularly in balancing the expected gains in terms of information about the intrusion of escaped farmed salmon in rivers.