Escaped farmed fish is one of the Norwegian aquaculture industry's biggest environmental challenges. Farmed salmon that migrate up rivers to breed may undermine the population's genetic material and resulting in a weakened population of wild salmon.

Simulations carried out by NINA with 20% escaped farmed salmon at spawning (close to the average in Norwegian rivers the last fifteen years) shows that there will be significant changes in the wild salmon stocks during ten salmon generations (about 40 years). It does not appear that farmed salmon establishes itself in rivers with a low number of escaped farmed salmon in the spawning population. But in rivers with a high number of escaped farmed salmon it appears that the population is gradually dominated by the offspring of farmed and hybrids of salmon. Even after many decades without new escapes, it is possible that these populations will be dominated by descendants of escaped farmed fish. NINA suggests that the average mix of escaped farmed in the spawning stocks should be less then 5%. An alternative value can that the gene flow from escaped farmed to wild salmon should be less than what is typically found between two wild salmon populations. Regardless, it is required that measures to sharply reduces the number of escaped farmed in nature must be implemented as possible.

Genetic interactions

There has long been concern that escaped farmed salmon may harm the various wild fish populations through hybridisation and altering the gene pools of wild populations (Hansen et al. 1991).There are several problems that can arise in this connection. If the farmed salmon have different characteristics and adaptations from wild salmon populations, gene flow may cause the wild salmon populations to lose characteristics that are crucial in a natural environment, while they adopt more of the farmed salmon's characteristics. On the other hand, if the escaped farmed salmon have less genetic variation than wild stocks, gene flow to the wild population will cause individual populations to lose variation (Tufto & Hindar). Variation is essential for two reasons (Hedrick, 2000), evaluated from both a short-term and a long-term perspective. A population that loses variation and thus becomes genetically uniform will be less resistant to disease and parasites. Or put another way: it is easier for a parasite to adapt to a population of genetically similar individuals (few polymorphic loci in the population and low heterozygosity) and where the individuals themselves have little variation (the individuals have few heterozygous loci). Additionally, in theory some of the harmful, recessive alleles will increase in frequency and produce less viable individuals (inbreeding depression). Studies just out (Reed & Frankham, 2003) empirically show that there is a good connection between fitness and heterozygosity, population size and quantitative genetic variation. Heterozygosity explains about 20% of the variation in fitness. In the long term, a population with little polymorphism will not have as great an evolutionary potential as a population with a lot of genetic variation.

All escaped farmed fish will come from a small number of farmed populations, which will lead to different populations becoming more like one another. It has also been claimed that coadapted gene complexes may dissolve. The following is an attempt to clarify relevant concepts and summarise empirical studies.

Evolutionary forces: mutation, selection, migration and drift

From modern evolutionary theory we know that there are four key forces that induce a population to change over time. These forces are selection, mutation, drift and migration. Even if all alleles originate in mutations, such events are all too rare to be an important force in ordinary evolution. It is therefore disregarded as a cause for fixing an allele or trait in a population. Drift is the random selection of gametes with different sets of alleles. The potential for evolution caused by drift is therefore inversely proportional to population size. Selection is probably the best-known evolutionary force and is caused either by survival and reproduction ("natural selection"), breeding ("artificial selection") or sexual selection. Migration (gene flow) means that individuals from a donor population reproduce in a recipient population. Migration results in the recipient population becoming more like the donor population. Although the donor and recipient populations will at the outset normally differ from each other in several characteristics and genes, unlike drift, migration will necessarily impact all these characteristics simultaneously. Like selection, migration is both deterministic and directional.1 For its part, drift is deterministic, but not directional. Selection can counteract migration, drift and mutations; if drift or migration has increased the frequency of alleles that produce low rates of survival and reproduction in earlier generations, natural selection can reduce them. Nevertheless, modern evolutionary theory says that migration is a very important evolutionary force, which can override selection and lead to less adapted populations (Graur & Li, 2000;Tufto, 2001; Lenormand, 2002).

Breeding and evolution

Traditional livestock breeding and evolution in the wild have several similarities. The biggest differences are that natural and sexual selection are more important in the wild, whereas artificial selection is more important in breeding programmes. In addition one can say that unlike natural selection, artificial selection has a goal. The fact that these mechanisms have certain similarities does not mean, as we try to elucidate in this chapter, that bred organisms are necessarily "natural" or harmless if introduced into the wild. Genetic drift is an important evolutionary force in small populations. A general rule of thumb2 says that populations of fewer than 500 individuals will lose genetic variation. After many generations, the genetic variation will no longer allow adaptations to the environment. A population of fewer than 50 individuals will after a few generations suffer from inbreeding depression.

Practically all traits have a genetic component. In addition, many traits have an environmental component, which is not inherited in the same manner. In an evolutionary perspective, only the genetic component is interesting. A trait may be physiological, behavioural, anatomical etc, and one or more genes may be controlling the trait. Most genes have a specific location in the genome. This place is called a locus. When a mutation occurs, a new variant of the gene arises. Such gene variants on the same locus are called alleles. All alleles once arose by a mutation, but mutations are so rare that they cannot be an evolutionary force. Many of the alleles can be found in a population's overall "genetic library", its gene pool.

Gene flow from farmed salmon to wild salmon

With a higher than 95% probability, wild salmon will return to the river they grew up in. The probability of migrating to the wrong river is greatest for geographically close rivers (Bentsen, 2000).This means that the salmon populations along the Norwegian coast are structured essentially according to what in ecology is called a "steppingstone model" (Kimura & Weiss, 1964.).

Explanation of the figure:
A model for gene flow in salmon. The circles represent the gene pools of the “population” of farmed salmon and of wild populations (B, C, D and E). The arrows represent gene flow between the populations, m is the migration rate between two geographically near salmon populations, m* is gene flow from the farmed fish population to the wild populations. Note that the gene flow from the farmed population to the wild populations is one-way, while there is gene flow both ways between two close wild salmon populations. The farmed fish are not affected by wild stocks, whereas the wild salmon populations affect one another. For Norwegian salmon, m < 0,05 is a likely estimate. For many wild populations, m* will be far greater, some places 20-30% (Fiske et al., 2000; Fleming et al. 2000). The effective population size (Ne) for farmed salmon would be from 30 to 125 (Bentsen, 2000). For wild salmon populations, it will normally be greater and much greater in the major salmon rivers.  

In the section "ecological impacts" we saw that the success in spawning and the life-long fitness of escaped farmed fish are reduced compared with wild fish (Fleming et al., 2000). Nevertheless, this yields a migration rate (m) equal to 0.19, which shows that there is a not unsubstantial gene flow to the wild population, even though the farmed salmon are poorly adapted to natural conditions. With a ratio of farmed salmon of about 50%, which was the case in this experiment, the genetic dissimilarities between the farmed population and the wild population was cut in half in 3.3 generations (Fleming et al., 2000).

Hybridisation with other species

Norwegian salmonids are good biological species with well-developed pre- and postzygotic barriers where they coexist. Although prezygotic barriers are greatest where the various species live together naturally (Youngson et al., 1993; Verspoor, 1988), hybrids may appear that are not completely sterile.This means that gene flow between species (introgression) can occur. However, the percentage of hybrids is very small, especially where the species live together naturally as they have done in Norway since the ice receded at the end of the last Ice Age. There is reason to believe that there is no extensive gene flow between wild populations of the species salmon (Salmo salar) and trout (Salmo trutta) in Norway. However, it has been shown that escaped farmed fish hybridise with trout to a greater extent than wild fish do: in individual stocks with a lot of farmed fish up to 7.5% hybrids between salmon and trout has been recorded (Hindar & Balstad, 1994; Youngson et al., 1993; 2001). In the short run this will lead to a decline in the stocks, while the long-range consequences of increased introgression are impossible to predict.

The breeding programme behind Norwegian farmed salmon

The basis population of the Norwegian breeding programme was taken from Norwegian salmon rivers in 1971 - 1974 (Gjedrem et al., 1991) and consisted of fertilised salmon ova from forty Norwegian rivers (Gjøen & Bentsen, 1997). Approximately twelve "full-siblinggroups" were taken from each river (Gjøen & Bentsen, 1997).The material gathered was divided into four lines where the fish were only crossbred internally.

Beginning in 1975, artificial selection for increased weight began by using the trait "slaughter weight" (Gjøen & Bentsen, 1997). Subsequently other traits were added, and (as of 2001), five traits are selected for (see Table 1). Experiments have also begun to investigate the genetic component to resistance against salmon lice (Soppeland, 2002), the purpose of which is to investigate possible heritability for eventual inclusion in the breeding programme.

Ne for the Aqua Gen lines where such data are available is between 30 and 40 (maximum 125) (Bentsen, 2000). This is far below what is recommended for the longterm maintenance of a population (Ne = 500). It is also below what is recommended for short-term preservation (Ne = 50). Of course, these are not hard and fast rules. Further, it is reasonable to believe that Ne is greater for the Aqua Gen breeding programme than it is for the private breeding lines. The best way to find answers to whether the breeding system is resulting in the loss of genetic variation is to investigate assumed neutral loci by using molecular markers. If there is less heterozygosity in the private breeding lines, this type of genetic contamination will be greatest from these lines.

Reduction in genetic variation

Mjølnerød et al. (1997) compared the genetic variation in neutral markers for wild salmon from Tana, Numedalslågen and one of the four farmed breeding lines from Aqua Gen. They used three types of genetic markers that are well suited for studying genetic variation at the population level: twelve polymorphic allozymes, three single-locus minisatellites and one multilocus minisatellite. The comparison showed that the number of alleles had been reduced for all the markers in the farmed breeding line. In addition, several of the allele frequencies diverged from what was expected. In sum, this indicates that drift has greatly impacted the farmed salmon's genetic material and that the farmed salmon have lost genetic variation as a result, which is supported by other literature in the field (e.g. Norris, et al., 1999). Loss of genetic variation has several unfortunate consequences; see 3.1.

Physiological/anatomical changes in farmed fish

Fleming & Einum (1997)5 showed that the farmed fish from Sundalsøra's population 1 (sensu Gjedrem et al., 1991), which has the preponderance of its genetic material from the river Namsen, displayed many anatomical differences compared with wild fish from the river Namsen. Fin lengths were shorter in the farmed fish, whereas the body form was more robust and less streamlined. The studies of Johnsson et al. (2001) showed that the heart rates of farmed salmon have become lower on average than in wild fish, and that these fish had poorer endurance and a weakened ability to flee from predators and migrate to suitable spawning beds. Furthermore, Johnsson et al. (1996) revealed that the production of growth hormones is higher in farmed salmon than in wild salmon. Besides the behavioural changes, this is probably one of the reasons for the rapid growth of farmed salmon (Fleming et al., 2000). Such characteristics may be a good adaptation in smolt facilities and cages, but are in all probability maladaptations in the wild. Singer et al. (2002) found that the smolt's ability to react/adjust in the transition from fresh water to salt water was poorer in farmed fish than in wild fish.

In this study, fish from population 1 (sensu Gjedrem et al., 1991) were compared with wild fish from the Imsa. The reason for the farmed fish's relatively poorer ability to adjust may therefore be either the result of the breeding programme or the expression of characteristics in the base population of the farmed fish. However, the reason is not so important, since regardless of the origin of the alleles, this will reduce the fitness of wild populations in the event of hybridisation.

Coadapted gene complexes

Many phenotypic traits are controlled by more than one locus. This implies that adaptations are founded on genes in more than one place in the genome. For such traits, selection has selected for allele combinations rather than alleles. Such selection works slowly and will not be able to counteract as high migration rates as single-locus traits can. Although to Bellona's knowledge, no studies have been published that have been able to identify this problem in Atlantic salmon (Salmo salar) in light of escaped farmed salmon, recent studies from Ireland (for the time being unpublished) point in this direction. Nonetheless we have strong circumstantial evidence from a study of a species of Pacific salmon (Oncorhynchus gorbuscha). This Pacific salmon species has a strict two-year life cycle. This means that in all rivers there are actually two populations, an odd-year and an even-year population that are temporally isolated and therefore never spawn during the same season. Such population pairs will necessarily be adapted to identical conditions, whereas the gene flow between them is minimal.

Gharrett et al. (1999) and Gharrett & Smoker (1991) cryopreserved milt and fertilised roe of the sister population the following year. The first hybrid generation evinced a good re-catch rate, but at the same time showed greater variation in size. The second-generation hybrids showed a low return migration rate. In general, it is not unusual for first-generation hybrids to show good or increased fitness. The reason for this is often that these individuals have a very high percentage of heterozygous loci. It is not until the second hybrid generation that the problems begin to become evident. The most obvious interpretation of this study is that there is more than one gene controlling many of the fitness-related traits in this species, in addition to the fact that the odd-year and even-year generations had coadapted gene complexes that were broken up in the second hybrid generation. This species is a relative of Atlantic salmon, and there is nothing that suggests that Atlantic salmon have genetics that are substantially different from Pacific salmon in this area. However, any difference may be due to different evolutionary histories, especially migration rates and how long the populations have been isolated.

Quantified effects of escaped farmed salmon in the river systems

It’s been conducted two experiments that quantify the impact of escaped farmed salmon in natural river systems. These are conducted in the river Imsa in Norway and Burrishole in Ireland.

The Imsa experiment started with releasing mature local salmon and farmed salmon upstream of a fish trap. The farmed salmon reproductive success was measured to be 16 percent of the wild salmon for one generation. Offspring of farmed salmon and hybrids had a faster growth rate than wild parr the first years. In addition, the total production was measured to be 30 percent lower than expected. This appears to be caused by negative interactions between farmed and wild salmon, in addition to a generally lower productivity in farmed salmon (Flemming, etc. 2000)

In Burrishole first and second generation offspring of wild and farmed salmon was compared with the three cohort roe that released upstream on a fish trap. The offspring of farmed and hybrids had poorer survival in nature. But they grew faster than the wild salmon witch got displaced in the younger stages of life.

Norwegian Institute for Nature Research (NINA) has constructed a model witch attempts to quantify the impact of escaped farmed salmon on wild populations of salmon (Hindar etc. 2007). The model calculations show that under the given conditions, in a population that receives 20 percent farmed salmon in each spawning period the genetic composition significantly changes during 40 years. The proportion of escaped farmed salmon in the spawning stocks in Norwegian rivers ranged from under 2 percent to more than 70 percent during the 90s. The report concludes that the amount of escaped farmed salmon in the spawning stocks should be kept below 5 percent.