The major threats to Atlantic salmon in Norway

Mechanisms and definition of terms

We define anthropogenic losses as reduction in the number of returning spawners due to the impact factors in freshwater or at sea caused by human activities. Anthropogenic factors can reduce Atlantic salmon populations through several mechanisms acting on different life stages. In freshwater, the carrying capacity may be reduced by decreased area or quality of the rearing habitat, ultimately causing reduced smolt production (Einum and Nislow, 2011). Several factors may increase juvenile mortality, and the effects of juvenile mortality on smolt production depend on when the mortality occurs relative to the strength and timing of density dependent population regulation (Milner et al., 2003^; Einum and Nislow, 2011). In the absence of marine fisheries, mortality of post-smolts at sea will proportionally reduce the number of returning adults to natal rivers (Atlantic salmon show precise homing to their natal rivers, Stabell, 1984^; Jonsson et al., 2003), because there appears to be no density regulation at sea (Jonsson et al., 1998).

We also consider the genetic integrity of populations in the classification system. Due to the extent of genetic structuring of Atlantic salmon, and evidence for genetically based adaptations among populations, conservation and management at the population level are recommended to protect diversity (Garcia de Leaniz et al., 2007^; Verspoor et al., 2007). The genetic integrity of a population may be threatened in several ways (e.g. interspecific hybridization, outbreeding due to immigration, inbreeding, and genetic drift due to reduced population size). The major impact on Norwegian Atlantic salmon populations is introgression (Allendorf et al., 2001) from escaped farmed Atlantic salmon (Glover et al., 2012^, 2013^; Karlsson et al., 2016).

We define a population as critically endangered or lost in nature when the number of returning adults is on average reduced by >75% over one generation (set to 5 years), or when significant introgression of farmed Atlantic salmon is documented (Glover et al., 2012^, 2013^; Karlsson et al., 2016) and >10% (Ryman et al., 1995^; Karlsson et al., 2016). We use the term “lost in nature”, because several Norwegian populations are conserved in the National Gene Banks (i.e. live fish kept in designated hatcheries and cryopreserved sperm; Bergan et al., 1991).

Anthropogenic factors considered

Fifteen anthropogenic impact factors were considered in the ranking (see below and Table 1). Reduced growth and survival at sea seem to contribute to declines in Atlantic salmon over large parts of the distribution range (Chaput, 2012). Feeding conditions at sea was not included in the ranking, because we lack evidence to support that this should be regarded as an anthropogenic impact factor in Atlantic salmon (Brander, 2007^; Friedland et al., 2009^; Rikardsen and Dempson, 2011). However, correlations between Atlantic salmon productivity and ocean climate indexes are documented (e.g. Friedland, 1998^; Friedland et al., 1998^, 2000, 2009^; Dickson and Turrell, 2000^; Jonsson and Jonsson, 2004^; Boylan and Adams, 2006), and this was considered in the ranking of climate change.

Open multimedia modal

Characteristics considered for each impact factor and scoring

Several characteristics were used to describe each impact factor along the two axes (Table 1). For each factor, the characteristics were scored from one to four based on quantitative or qualitative criteria (Table 1), and thereafter scores were added up and expressed as a proportion of the potential maximum score. In a few cases, half points were given.

For the effect axis, five characteristics were considered for each impact factor. The number of affected populations was used as a measure of the extent of each factor. The geographical distribution was also considered, because impact factors operating on a regional rather than local level may promote loss of regional ecotypes or genotypes. The criteria ranged from effects on few dispersed populations to effects on a national level (affecting populations in at least 14 of the 16 counties with Atlantic salmon populations in Norway; Figure 1). The typical effect on affected populations was assessed in terms of reduction in number of adult returns, based on knowledge from published scientific papers and technical reports. The criteria ranged from small reductions (<10%) in adult returns to very large reductions (>75%). It can be difficult to characterize a “typical” effect, but classification into one of four relatively wide groups should be robust. The number of critically endangered or lost populations in nature was quantified based on the Norwegian Environment Agency database (Norwegian Environment Agency, 2016b), where lost populations (no longer self-reproducing) are listed. Critically endangered populations in terms of genetic integrity was added to these based on published studies (Skaala et al., 2006^; Glover et al., 2012^, 2013^; Karlsson et al., 2016). The criteria ranged from no affected populations to more than 20 affected populations. Finally, implemented mitigation measures that have reduced loss, or the likelihood of further populations becoming critically endangered or lost, were considered. Some of the impact factors have been active for many years and several mitigation measures have been implemented, such as the eradication programme against the introduced parasite Gyrodactylus salaris and large-scale liming in acidified watercourses (see below). Here, we assessed the documented effects of the measures, based on published papers, Norwegian reports and databases. The classification ranged from extensive measures with good effects to few or no measures or measures with no net effect in terms of reducing the impact. The extent of knowledge of each of the impact factors was classified as extensive, moderate or poor, based on assessment of published scientific papers and technical reports.

For the development axis, three characteristics were considered for each impact factor. The potential for effective measures to be implemented was based on a projection of the present situation and mitigation measures, primarily based on expert judgement of public information such as action plans, governmental White Papers and regulations and guidelines from the relevant management bodies, and general knowledge of primary experts within the Atlantic salmon committee. Classification ranged from few or no effective measures being planned to preparation of extensive and effective mitigation measures. Next, the likelihood of further losses in the number of returning adults and the likelihood of additional populations becoming critically endangered or lost were ranked from low to very high, based on expert judgement of public documents from the government departments responsible for each impact factor. The uncertainty of the projected development of each of the impact factors was classified as small, moderate, or high (scores 1–3).

Procedure for factor scoring

One or more primary experts from the Atlantic Salmon Committee presented initial scoring for their assigned impact factor at a committee meeting. The presentation included relevant scientific articles, Norwegian reports and other public documents. Scoring was extensively discussed within the committee, before arriving at the final scoring. The assessment of the primary experts was emphasized in the discussions. Each member could present alternative assessments and scores in the annual report if they did not agree in the scoring made by the committee majority, but this never happened because consensus was always reached.

Hydropower regulation

Hydropower regulation includes hydropower facilities such as dams and power stations, but also altered water discharges and temperatures during the year due to water detraction or storage of water in reservoirs. The effects on Atlantic salmon vary substantially among rivers (Johnsen et al., 2011^; Birkel et al., 2014). This impact factor ranked high along the effect axis because it affects nearly 20% of the populations, and 19 populations may have been lost due to hydropower development (Johnsen et al., 2011). The effect in terms of reduced adult returns was classified as being between moderate and strong, based on published studies from Norway and elsewhere (Murchie et al., 2008^; Ugedal et al., 2008^; Johnsen et al., 2010^, 2011^; Otero et al., 2011^; Birkel et al., 2014^; Hvidsten et al., 2015^; Stich et al., 2015^; Bilotta et al., 2016). Hydropower developments in Norway started in 1882, peaked during 1945–1961, and many of the extensive regulations were completed by 1980 (Johnsen et al., 2011). Since then, few hydropower stations have been built in Atlantic salmon rivers. In two White Papers, the government designated 29 fjords distributed along the entire coastline as “national Atlantic salmon fjords” and 52 rivers draining into these fjords as “national Atlantic salmon rivers”, to protect the wild Atlantic salmon (Anon., 2001–2002, 2006–2007). These rivers, representing 72% of the conservation limits (CLs) for Atlantic salmon in Norway, were given protection against further hydropower development, water abstractions and flood control measures. There has been a general decrease in the proportion of hydropower development schemes that have been granted concessions or permissions after 2003, and this was particularly evident in the national Atlantic salmon rivers (Vøllestad et al., 2014).

Implementation of the European Water Framework Directive (WFD, EU, 2000) is also expected to cause improvements in the management of regulated rivers. Moreover, in a joint report from the Norwegian Environment Agency and the Norwegian Water Resources and Energy Directorate, 35% of the regulated watercourses with anadromous salmonids were given top priority in a national plan for revision of hydropower concessions (Sørensen et al., 2013). Such revisions will likely result in improved conditions for Atlantic salmon due to inclusion of minimum flow requirements and habitat restoration (Forseth and Harby, 2014). The regulatory framework (national Atlantic salmon rivers, WFD and revisions) and implemented mitigation measures resulted in low scores along the development axis. Increased hydropeaking caused by rapid increase or decrease in water discharge in accordance with fluctuations in the power demand (Harby and Noack, 2013), cumulative effects of small hydropower projects in tributaries (Benejam et al., 2016) and the increasing demand for renewable energy to reduce carbon emissions (e.g. the EU Renewable Energy Directive, EU, 2009) add uncertainty to future development. Moreover, a recent White Paper signals that the Norwegian Government in some cases may allow for new hydropower developments in protected rivers, particularly in relation to flood protection (Anon., 2015–2016).

Water abstraction

Water abstraction in salmon rivers includes the use of water for fish hatcheries, industry and irrigation. Abstraction for irrigation purposes is limited due to a climate with high annual precipitation, and industrial water abstraction from rivers is not common in Norway. As a result, few rivers are severely impacted, and impacted rivers are often small and dominated by brown trout (Salmo trutta L.). A few populations have been lost, at least partly due to water abstraction for hatcheries, but the typical effects were regarded moderate. The low number of affected populations caused water abstraction to score low along the effect axis. Increased awareness through recent impact studies (Bergan, 2014) and implementation of the WFD reduce future risk of populations being critically reduced by water abstraction. Therefore, water abstraction scored medium along the development axis. However, increased smolt production for aquaculture may increase the demand for further water abstraction (Kittelsen et al., 2006) and adds uncertainty to future development.

Habitat alterations

Man-made structures and activities such as channelization, dredging, and building of levees and impoundments such as weirs lead to changes in river morphology and hydrology, which often lead to reduced habitat quality for fish, or even lost habitat (Kemp, 2010). Such habitat alterations are the results of flood protection or land reclamation and are common in Norwegian rivers (>200 rivers affected). However, extensive alterations are rare in Norway and typical effects were described as small. In summary, physical habitat alterations scored relatively high along the effect axis. However, on the development axis, the score was low, reflecting the river habitat protection offered by the Watercourse and Groundwater Act (2000) and implementation of the WFD, as well as several recent measures and plans to improve salmon habitats (Fjeldstad et al., 2012^; Hauer et al., 2015). In western Norway, several recent major flooding events have resulted in new flood protection plans, including channelization, dredging and levees. These measures may challenge the protection of Atlantic salmon habitat.

Acidification

Deposition of long-range transported air pollutants (sulphur and nitric acids; acid rain) in areas of bedrock with low neutralizing capacity results in lowered pH and release of potentially toxic aluminium (Al) to surface waters (Driscoll et al., 1980). This combined effect (acidification) has affected salmon populations in large areas in southern Norway (Hesthagen and Hansen, 1991^; Rosseland and Kroglund, 2011). Increased fish mortality may be related to both high concentrations of H⁺ ions (reduced pH) and inorganic monomeric Al (Rosseland and Staurnes, 1994^; Gensemer and Playle, 1999). Interaction of Al with the ion exchange and oxygen transport systems across the gills of juveniles and smolts accounts for the major part of the mortality during these life stages. Sub-lethal effects in freshwater may be lethal when the fish enter seawater, causing increased post-smolt mortality (Kroglund et al., 2007^; Thorstad et al., 2013). Large reductions in salmon catches due to acidification were already recorded around 1900 (Hesthagen and Hansen, 1991). Fish mortality and population losses increased with the increase in sulphur emissions during 1960–1980. By 1990, Atlantic salmon populations were lost in 25 rivers and greatly reduced in at least eight more rivers (Hesthagen and Hansen, 1991). Consequently, acidification scored high along the effect axis (typical effects described as very large and many populations have been lost), but the regional distribution and successful measures (see below) prevented top scores. From about 1990, significant reductions in acid deposition has resulted in increased surface water pH and reduced concentrations of Al in southern Norway (Skjelkvåle et al., 2003). This chemical recovery has not been sufficient to create acceptable water quality for Atlantic salmon.

A national liming programme was started in 1983 (Clair and Hindar, 2005), and 22 salmon rivers are limed by continuous addition of limestone powder. New salmon populations have established in affected rivers by re-stocking and straying of salmon from other rivers (Hesthagen et al., 2011). The national monitoring programme has documented only minor reductions in acid deposition since 2010, and further reductions will probably be negligible. Consequently, no significant change in this threat is expected in coming years. Moreover, as long as the liming programme continues (Norwegian Environment Agency, 2016a), no further losses of salmon populations are expected. Recently, increased funding has allowed optimization of the liming programme in some rivers and expansion of the programme by one or two rivers from 2016 to 2017. The likelihood of further losses was thus classified as low.

Gyrodactylus salaris

The ectoparasite G. salaris was accidently introduced to Norway during the 1970s, with devastating effects on infected Atlantic salmon populations (Johnsen and Jensen, 1991). Juvenile densities were typically reduced by 90% or more, and adult returns plummeted in affected watersheds (Johnsen and Jensen, 1991). In total, 50 Norwegian Atlantic salmon populations distributed in eight fjord systems have been infected. The parasite has limited survival in saline water (Soleng and Bakke, 1997), but within each fjord system, G. salaris can disperse to new populations by infested fish moving within brackish water (Soleng et al., 1998). However, dispersal to new fjord systems across steep salinity gradients requires other transmission pathways (e.g. transport of fish or water). No other impact factor has caused more populations to become critically endangered or lost in Norway, and consequently it scored high along the effect axis. Various eradication programmes have been in operation, mainly using the toxin rotenone to kill all the fish in the impacted watersheds, because the parasite do not survive without its long-term hosts. Moreover, artificial barriers to reduce the area for chemical treatment and recently Al sulphate treatment (Soleng et al., 1999^; Pettersen et al., 2006) have been used. Eradication efforts increased substantially after revisions of the National Action Plan against G. salaris in 2008 and formulation of the current plan in 2014 (Norwegian Environment Agency, 2014). Consequently, by the end of 2015, G. salaris was present in 7 populations, 22 populations were declared free from the parasite, and 21 populations were in the process of being declared free of the parasite (requiring a minimum of five years without parasite observations). The action plan contains specific plans for removing the parasite from four of the remaining populations, whereas an expert group appointed by the Norwegian Environment Agency is currently exploring treatment options for the last three populations. The Atlantic salmon populations have been re-established from the National Gene Bank for Atlantic salmon (Bergan et al., 1991) after eradication of the parasite. Thus, measures taken to reduce the effect of the parasite have been extensive and largely successful. Although the parasite has recently re-emerged in one population and been found in some small river systems within infected fjord systems, no dispersal to new fjord areas have been documented since 1996. The probability of further losses is thus moderate, the action plan outlines extensive future measures to reduce this threat, and consequently it scored low along the development axis.

Salmon lice

The salmon louse (Lepeophtheirus salmonis) is an ectoparasite of salmonids in the sea. Historically, salmon lice were observed in moderate numbers on wild salmonids, but because also farmed Atlantic salmon act as hosts, open net cage farming has increased the production of salmon lice in many coastal areas (Finstad and Bjørn, 2011^; Thorstad et al., 2015). Since the late 1980s, salmon lice epizootics have been reported in wild salmonids in Norway, Scotland, Ireland, and Canada (Heuch et al., 2003^, 2005^; Revie et al., 2009^; Finstad et al., 2011^; Thorstad et al., 2015). In farm-intensive areas, lice levels on wild salmonids are typically higher, and more variable than in farm-free areas (e.g. Serra-Llinares et al., 2014^, 2016^; Helland et al., 2015^; Thorstad et al., 2015). The individual physiological and pathological effects of salmon lice on salmonids have been thoroughly described (Finstad and Bjørn, 2011), and population level effects have been well documented for Atlantic salmon. Studies in Norway, Scotland, and Ireland, based on comparison between chemically protected post-smolt and untreated controls, have shown average additional marine mortalities by salmon lice of 0.6–39% across locations and years (Gargan et al., 2012^; Jackson et al., 2013^; Krkošek et al., 2013^; Skilbrei et al., 2013^; Vollset et al., 2016). Salmon lice may also increase age at maturity (Vollset et al., 2014). Hence, the effects in terms of reductions in the number of returning adults may be substantial, and typical effects in affected populations were classified as large (>25%). Salmon lice may reduce marine survival of wild Atlantic salmon in farmed areas along the Norwegian coast, especially in the southwestern and middle parts, but also in parts of northern Norway (Taranger et al., 2015^; Svåsand et al., 2016). Thus, the number of affected populations is high, and the geographical distribution was classified as regional to national. Large effects on many populations gave high scores along the effect axis. The efforts to reduce salmon lice infestation pressures on Atlantic salmon are substantial (Norwegian Food Safety Authority, 2016), and during the last 30 years, infestation pressures have varied with the size of the aquaculture industry and the implemented measures. However, the large growth of the industry has repeatedly nullified the effect of the measures (Heuch and Mo, 2001). Risk assessments of the environmental impacts of salmon farming (Taranger et al., 2015^, Svåsand et al., 2016) have shown a general increase in infection pressure of salmon lice and its geographical distribution from 2010 to 2015. Moreover, resistance among salmon lice to the major drugs used to control lice levels in the farms is a major problem (Grøntvedt et al., 2016). Warming of coastal waters (Hoegh-Guldberg and Bruno, 2010) may extend the problem to the northernmost part of Norway. Consequently, the probability of further losses in number of returning Atlantic salmon was classified as high. High infection pressures over years may also reduce the number of returning adults to the extent that populations become critically endangered, and this probability was classified as moderate to high. A White Paper from the Ministry of Trade, Industry and Fisheries outlines further growth of the farming industry but also new instruments and management principles to ensure sustainable growth of salmon farming (Anon., 2014–2015).

Infections related to fish farming

Wild fish are the original source of pathogens causing diseases in farmed fish, but farming provides conditions for proliferation and spread among farmed fish and back to wild fish (Johansen et al., 2011). Moreover, farming practices have introduced new pathogens to naïve populations (Peeler et al., 2011). The use of fish from Scotland and Denmark in aquaculture operations in Norway provided a vector for the transmission of the bacterium Aeromonas salmonicida subsp. salmonicida, causing furunculosis in wild Atlantic salmon populations (Johansen et al., 2011). In 2013, outbreaks of one or more viral diseases were registered in 38% of the c. 600 active farming localities, and a large number of farms had outbreaks of bacterial and parasitic diseases (Bornø and Lie Linaker, 2015). Hence, the potential infection pressure from farmed to wild Atlantic salmon is large. The knowledge on pathogen transfer from farmed to wild Atlantic salmon is poor (except regarding salmon louse), partly due to the challenges of documenting the transfer and disease outbreaks in wild fish (Johansen et al., 2011). However, using Piscine orthoreovirus (PRV) as a model, Garseth et al. (2013) presented a strong case for virus transmission from farmed to wild salmonids in Norway. Moreover, escaped Atlantic salmon infected with both salmon alphavirus and PRV have been found to ascend a river close to the likely source farms (Madhun et al., 2015). There is no studies documenting large effects in wild Atlantic salmon populations by infections originating from farmed salmon, and the typical effect was classified as small. However, this is an area where knowledge is largely lacking. Given the number of infective organisms (i.e. virus, bacteria, fungi, and parasites), and their effects in farmed fish, there is potential for strong effects in wild populations. Even though the typical effect was classified as small, increased knowledge through future studies may change this classification. The number of populations potentially affected is high, and the measures are likely unable to reduce the likelihood of further losses.

Infections linked to other anthropogenic activities than fish farming

For several infective organisms, disease outbreaks in Atlantic salmon can be related to changes in environmental conditions caused by other anthropogenic activities than fish farming. An outbreak of proliferative kidney disease (PKD) was linked to reduced flow and increased temperatures due to hydropower regulation, and the juvenile mortality was substantial (Sterud et al., 2007). The parasite Tetracapsuloides bryosalmonae causing PKD has been found in several other Norwegian rivers (Mo and Jørgensen, 2017). Furunculosis outbreaks may also occur under similar environmental conditions, with massive deaths of adult fish (Johnsen and Jensen, 2005). Due to the lack of a comprehensive national monitoring programme, the knowledge on this impact factor is poor and assessment was mainly based on expert judgement. The impact was ranked as moderate along both axes. Wide geographical distribution, the lack of effective measures, predicted increases in summer temperatures, and reductions in summer flow due to climate change (Schneider et al., 2013) are important for the classification.

Escaped farmed Atlantic salmon

River scale experiments in Imsa and the Guddalselva in Norway and Burrishoole in Ireland have documented how escaped farmed Atlantic salmon, their offspring and hybrid offspring from mating with wild Atlantic salmon can affect wild populations negatively (Fleming et al., 2000^; McGinnity et al., 2003^; Skaala et al., 2012). Introgression of farmed Atlantic salmon into wild populations may result in lower adult returns due to reduced smolt production (Fleming et al., 2000) and reduced sea survival (McGinnity et al., 2003). The effects from introgression of farmed Atlantic salmon on a given population may vary, because the effects depends on extent of local adaptation to environmental factors, which may differ among populations (Fraser et al., 2010a). Loss of genetic integrity of wild populations due to introgression from farmed Atlantic salmon is a fundamental threat. Farmed Atlantic salmon have been through strong domestication selection and differ from wild Atlantic salmon in a number of genetically based traits (e.g. growth rate: Thodesen et al., 1999^; Glover et al., 2006^, 2009^; Solberg et al., 2013a^, b, stress tolerance: Solberg et al., 2013a; behaviour: Einum and Fleming, 1997), potentially leading to maladaptive trait values for life in nature (Ferguson et al., 2007). Moreover, reduced genetic variability is documented in farmed Atlantic salmon strains, both due to the selection in the breeding programmes and the number of parents used (Norris et al., 1999^, Skaala et al., 2004, but see Rengmark et al., 2006). Repeated introgression of farmed Atlantic salmon into wild populations may over time cause the local Atlantic salmon populations to be replaced by less adapted and less genetically variable hybrid populations. Glover et al. (2012) showed genetic changes in microsatellite DNA in 6 of 21 studied populations (28%) that could be linked to introgression of farmed Atlantic salmon. In further studies using single nucleotide polymorphism (SNP) markers in 20 of the same populations, Glover et al. (2013) documented significant changes in five populations, and found that their present genetic profile was closer to a mixed farmed sample than historical samples from the population. The proportion of genes from farmed Atlantic salmon in the wild populations varied from 2 to 47%. Karlsson et al. (2016) recently documented that 51 of 109 Norwegian Atlantic salmon populations showed significant genetic introgression from farmed salmon (using diagnostic SNP marker developed by Karlsson et al., 2014). The mean introgression level in all rivers was 6.4%, and 27 populations (25%) had introgression levels above 10% (maximum 42%). Thus, we classified the number of critically endangered or lost populations due to this impact factor at the maximum level (>20 populations). Karlsson et al. (2016) found that introgression had occurred in all regions of Norway, but was highest in the regions with most farming. They also found a relationship between introgression levels and average proportion of escaped farmed salmon found in the rivers during monitoring (Fiske et al., 2006), although parts of the variation remain unexplained. The genetic studies and monitoring of farmed escapees in the rivers clearly show that escaped farmed Atlantic salmon is a national threat affecting a large proportion of the populations. The typical effect of introgression by farmed salmon in terms of reduced adult returns was classified as moderate.

The efforts to reduce the number of farmed Atlantic salmon escapes have been considerable, and the official number of reported escapees decreased from a peak level at more than 900 000 individuals in 2006 to generally below 300 000 thereafter (Norwegian Directorate of Fisheries, 2016a). Similarly, the incidence of farmed Atlantic salmon in samples from wild spawning populations has decreased from an average of 20–35% across monitored populations before 1998, to a level between 9 and 18% after 2003. However, since both the ecological effects (competition from farmed and hybrid offspring) and introgression are cumulative across generations (McGinnity et al., 2003^; Fraser et al., 2010a^, b), the measures taken are regarded as unable to reduce the likelihood of further losses.

Introduced fish species

Norway has 32 native self-sustaining freshwater fish species and at least 11 non-native self-sustaining species (Huitfeldt-Kaas, 1918^; Hesthagen and Sandlund, 2007). There is limited information on how introduced non-native and translocated native species affect Atlantic salmon, but they may potentially affect juvenile survival through competition for space and food. Some of the species may have significant impacts if they establish large populations.

Oncorhynchus-species use similar freshwater habitats as Atlantic salmon (Quinn, 2005). It is uncertain if rainbow trout (O. mykiss) has established viable populations in Norwegian rivers, even if some indications exist (Sægrov et al., 1996). Rainbow trout is produced in net pen aquaculture, of which some escape into the wild (reported average at c. 80 000, range: 200–315 000, during 2001–2015). Potential spawning of rainbow trout in spring may lead to excavation of newly hatched Atlantic salmon larvae. Competition between rainbow trout and Atlantic salmon parr may also be expected. Pink salmon (Oncorhynchus gorbuscha) have been stocked in in Russian rivers (Gordeeva and Salmenkova, 2011), leading to dispersal to Norwegian rivers, particularly in the north, where a few populations appear to have been established (Bjerknes, 1977^; Bjerknes and Vaag, 1980; County Governor of Finnmark pers. Comm.). Pink salmon often utilize spawning habitats close to the estuary, and juveniles migrate to the sea soon after hatching (Quinn, 2005). Thus, competition with Atlantic salmon juveniles will be minor. The ecological effects of the establishment of rainbow trout and pink salmon on Atlantic salmon are basically unknown (Quinn, 2005).

Other introduced species, such as gudgeon (Gobio gobio) (Eken and Borgstrøm, 1994), or translocated species such as pike (Esox lucius) and bullhead (Cottus gobio), may interact with Atlantic salmon through predation (Kekalainen et al., 2008) or competition (Gabler and Amundsen, 1999^; Jørgensen et al., 1999^; Tammi et al., 2003).

The effects of these introductions are poorly studied and therefore uncertain. Because the few studies that exist on introduced or translocated species in Norway indicate small effects (Gabler and Amundsen, 1999^; Jørgensen et al., 1999), this factor was ranked relatively low along the effect axis. There is uncertainty to what extent introduced or translocated species will further expand their ranges, particularly given the observed climate change. The extensive aquaculture industry will continue to produce rainbow trout that escape from net pens, leading to a large propagule pressure (Colautti et al., 2006). The likelihood of further losses was considered as moderate.

Agricultural pollution

Many Norwegian Atlantic salmon rivers run through valleys with agricultural activity, while large parts of their catchments are located in sparsely populated low productive areas. Only 3–4% of the area of Norway is farmland (Statistics Norway, 2016). Runoff of phosphorous (P) from agriculture may thus stimulate productivity in the generally nutrient-poor Norwegian salmon rivers (Jonsson et al., 2011^; Foldvik et al., 2016). Cultivation of new areas and runoff from exposed farmland (e.g. newly ploughed) may reduce Atlantic salmon habitat quality due to erosion and transport of fine particulate matters to the rivers (Kemp, 2010). Silage effluents and high stocking rate of grazing animals may result in high input of easily degradable organic matter and oxygen depletion in streams (Foy and Kirk, 1995), but salmonids in large rivers are probably less affected due to sufficient dilution and oxygen supply. Runoff of pesticides from treated areas is considered under “hazardous substances”.

Agricultural pollution potentially affects a number of Atlantic salmon populations and is a national impact factor. The National Action Plan against Agricultural Pollution for 1985–1988 developed mitigation strategies consisting of a set of legislative, regulatory, and economic instruments as well as information campaigns (Bechmann et al., 2008). Long-term monitoring in small streams in agricultural areas has showed variable temporal trends in nutrient losses (Bechmann et al., 2008^; Stålnacke et al., 2014), but a recent evaluation (Skarbøvik et al., 2014) concluded that implemented measures have positively influenced river water quality. The typical effect in salmon rivers was classified as small. Moderate nutrient loads stimulate productivity and high loads are rare due to relatively small agricultural areas in the catchments of most salmon rivers. No Atlantic salmon population has been classified as critically endangered or lost due to agricultural pollution (Norwegian Environment Agency, 2016b). Consequently, the rank along the effect axis is relatively low. Implementation of WFD management plans will likely initiate further countermeasures. This impact factor thus scored low also along the development axis. Increased runoff of nutrients and soil due to climate change (Deelstra et al., 2011) adds uncertainty to future development.

Hazardous substances

Atlantic salmon watercourses may receive heavy metals, pesticides, organic micropollutants and radionuclides from local (of natural and anthropogenic origin) and distant (long-range transported) sources (Rosseland and Kroglund, 2011). Norway is a rural country with a relatively small onshore industrial activity. Major industrial facilities are typically situated in the lowermost parts of rivers and at estuaries where dilution may be adequate to avoid harmful effects. Also, industry effluents are regulated by discharge permits from the Norwegian Environment Agency.

The effects on fish vary from sub-lethal effects to long-term reductions in survival (Rosseland and Kroglund, 2011) depending on the substance and exposure. Some pollutants (so-called hormone mimics) may influence the development of sex and gonads, with potentially strong effects on fish reproduction (e.g. Moore and Waring, 2001). The typical effects were considered moderate, but knowledge on effects on Atlantic salmon populations are limited. Two Atlantic salmon populations have been lost at least partly due to industrial pollution, but new populations have been established in both rivers.

The EU Commission has listed several of the most relevant substances as prioritized harmful substances in the WFD due to their potential toxic effects and has set limits for their concentrations in freshwaters and the sea. The aim is to phase out the use of these compounds. Thus, this impact factor scored relatively low along the development axis.

Mining

Mining for metals and minerals, together with quarries for production of different crushed bedrock products may affect surface water quality. All these activities produce particles that may be transported to rivers as suspended solids. The effects are poorly documented, but may be both direct (e.g. mechanical damage of gills) and indirect (e.g. clogging of spawning sites). Leaching of heavy metals (e.g. copper) from waste rock dumps and from flooded mines has probably the greatest potential for effects on salmon populations. Smolt is the most sensitive life stage, and smolts may be affected at very low concentrations of heavy metals (Rosseland and Kroglund, 2011). However, concentrations of heavy metals are generally below critical levels in Norwegian salmon rivers.

Potential effects depend on which minerals are mined, runoff treatment, and the downstream dilution. In Norway, harmful runoff from the mining industry has been related to abandoned mines, especially those based on blasting of sulphidic, metal-containing minerals. Sulphides are oxidized on exposure to air, producing sulphuric acid, potentially causing elevated concentrations of harmful Al and heavy metals such as copper, nickel, and zink. Sea deposits of waste rock, which are planned in Norway, may also influence Atlantic salmon negatively (Ramirez-Llodra et al., 2015), but studies from marine systems are lacking. Effects of mining on Atlantic salmon probably varies largely among sites, but was classified as small.

Overexploitation

After the establishment of CLs (in numbers of eggs or mass of females) based on stock-recruitment relationships (Hindar et al., 2011^; Forseth et al., 2013), overexploitation can be defined as reduction of the spawning population below the CL due to exploitation. Overexploitation is a dynamic impact factor, often with relatively rapid population responses to reduction in exploitation. Exploitation of Atlantic salmon was likely low until effective gears was developed during the 1800s (Shearer, 1992). In Norway, the exploitation level peaked during the 1970 and 1980s due to the large marine driftnet fishery (Jensen et al., 1999), but decreased after closure of this fishery in 1989 (Hindar et al., 2011), and has further decreased with recent reductions in other costal fisheries (Forseth et al., 2013). The Atlantic Salmon Committee estimates annually both attainment of the CLs (Forseth et al., 2013) and overexploitation (as a percentage of the CL) for 160–180 of the largest Atlantic salmon populations in Norway (representing over 90% of the annual total river catches). Management targeted at reaching CLs seccessfully reduced exploitation both in freshwater and along the coast, and improved attainment of the CLs (Forseth et al., 2013). By 2015, average overexploitation of assessed populations (weighted by the CLs, as a measure of population size) was 14%, and average attainment of CLs was 87%. This result was largely impacted by overexploitation in the Tana watercourse, where overexploitation is considerable, and management according to CL has not been fully implemented because management is regulated by a bilateral agreement between Finland and Norway. The Tana watercourse affects the national result due its large size (if all CLs were reached, size of the salmon populations in this watercourse alone would amount to one-fifth of the total size of all assessed populations). Omitting the Tana population complex, average overexploitation was 7.7%, and was found in 52 of the 190 assessed populations. Thus, overexploitation scored low along both axes, largely due to the recent management measures that have reduced exploitation.

Predation

Atlantic salmon are vulnerable to predatory birds, mammals, and fish (Ward and Hvidsten, 2011). This is generally not considered an anthropogenic factor, but is included to the extent that predation is influenced by human activities. Human activities may increase the number of predators (Ward and Hvidsten, 2011) and the vulnerability of salmon to predation. Overexploitation or habitat changes may reduce salmon abundance to the point that salmon become disproportionately vulnerable to predators (Ward et al., 2008). Human activities may also increase the exposure to predation. Dams and reservoirs may for instance slow down Atlantic salmon migration, create favourable habitats for predators, and concentrate predators and prey (e.g. Blackwell and Juanes, 1998^; Jepsen et al., 1998^; Aarestrup et al., 1999). River regulations may also cause loss of ice cover, resulting in increased exposure to ectothermic predators (Valdimarsson and Metcalfe, 1998).

There is little data available to assess how and to what extent predation affect salmon populations, but there is little doubt that predation influences behaviour, recruitment, and population dynamics of salmon (Ward and Hvidsten, 2011). In Norway, there are likely few Atlantic salmon populations impacted by predation influenced by human activity. Removal of ice cover after hydropower regulation in northern rivers is likely the main challenge. The challenge with smolt predation in hydropower reservoirs occurs in some rivers, but is generally rare in Norway. Based on available information and expert judgements, predation due to human activity was ranked low along both axes.

Climate change

Recent and projected climate changes represent major demographic and adaptive challenges to Atlantic salmon (Todd et al., 2011). Recent reviews conclude that populations have been and will be affected by climate change, both in freshwater, during seaward migration (Jonsson and Jonsson, 2009^, 2011^; Todd et al., 2011) and in the marine environment (Friedland et al., 2003^; Reist et al., 2006^; Jonsson and Jonsson, 2009^, 2011^; Todd et al., 2011^; Otero et al., 2014). Negative effects are particularly likely to occur in the southern distribution rage of Atlantic salmon (Todd et al., 2011), where physiological tolerance thresholds may be exceeded and summer droughts become more frequent (Bates et al., 2008^; Schneider et al., 2013). An individual based mechanistic population model (Hedger et al., 2012) was used to predict climate change effects on Atlantic salmon from three climatic regions of Norway (Hedger et al., 2013) based on predicted local stream temperatures and discharges obtained from downscaled global climate models. According to these models, increased summer temperatures under future climate regimes increased Atlantic salmon production in western and northern Norway, whereas reduced summer wetted area caused lower predicted smolt production in southern Norway. Climate change may thus have both positive and negative effects on the Atlantic salmon freshwater production. The model included the whole life cycle, but did not consider effects in the marine environment. Several studies show correlations between growth or survival and ocean temperature or climate indices (Jonsson and Jonsson, 2011^; Todd et al., 2011), and a relationship between growth and survival has been established (Friedland et al., 2000). However, the mechanistic relationships are unclear, and correlations may be due to direct physiological effects of temperature, or due to indirect changes in prey abundance or quality (Beaugrand and Reid, 2003^, 2012). The thermal scaling of Atlantic salmon growth at sea is poorly described in Forseth et al. (2011). Moreover, Atlantic salmon is a generalist and opportunistic predator at sea (Rikardsen and Dempson, 2011). Given the wide oceanic distribution and limited knowledge of migration routes of different populations (Dadswell et al., 2010), it is difficult to establish links between changes in prey abundance, growth, and survival. However, the general and continued decline in Atlantic salmon marine survival during the last decades (e.g. Friedland et al., 2003^; Jonsson and Jonsson, 2004^; McCarthy et al., 2008^; Chaput, 2012), suggest observed changes in ocean climate (Hoegh-Guldberg and Bruno, 2010) may be causative.

Considering both the freshwater and marine environments, the typical effect of climate change on Norwegian Atlantic salmon was classified as small, but the number of populations affected, the geographical distribution (regional) and the lack of effective measures caused moderate scores along both axes. However, current knowledge can be classified as poor and the uncertainty of future development is high.

The overall ranking of the threats to Norwegian Atlantic salmon

Four major groups of impact factors were identified according to the overall analysis of scores for each impact factor (Table 1, Figure 2). Escaped farmed Atlantic salmon and salmon lice were identified as expanding population threats with high scores along both axes. Escaped farmed Atlantic salmon had the highest scores along both axes, and was identified as the largest threat. The second group, G. salaris, acid rain, hydropower regulation and physical habitat alterations were also identified as population threats with high scores along the effect axis, but were classified as stabilized, mainly because effective mitigation measures have been implemented and there are plans for further mitigation measures in the near future.

The third group, positioned in the middle of the diagram (Figure 2), consisted of other infections related to fish farming, infections linked to other anthropogenic activities than fish farming, and climate change. The knowledge on the effects of these impact factors is generally poor and the uncertainty for the projected development particularly high.

The forth group, positioned towards the lower left corner of the diagram, represents stabilized loss factors. Among these, agricultural pollution ranked highest along the effect axis but low along the development axis. Water abstraction ranked highest in terms of risk of further losses. Overexploitation was ranked low both on the effect axis and development axis because of major reductions in fisheries due to implementation of management according to CLs (Forseth et al., 2013). Introduced fish species, mining, hazardous substances and predation were the other factors in this group.

Acknowledgements

We acknowledge the contribution of previous members of the Atlantic Salmon Committee, Kjetil Hindar, and Frode Kroglund, who participated in the development of the classification system.