Anadromy, potamodromy and residency in brown trout Salmo trutta: the role of genes and the environment

Abstract Brown trout Salmo trutta is endemic to Europe, western Asia and north‐western Africa; it is a prominent member of freshwater and coastal marine fish faunas. The species shows two resident (river‐resident, lake‐resident) and three main facultative migratory life histories (downstream–upstream within a river system, fluvial–adfluvial potamodromous; to and from a lake, lacustrine–adfluvial (inlet) or allacustrine (outlet) potamodromous; to and from the sea, anadromous). River‐residency v. migration is a balance between enhanced feeding and thus growth advantages of migration to a particular habitat v. the costs of potentially greater mortality and energy expenditure. Fluvial–adfluvial migration usually has less feeding improvement, but less mortality risk, than lacustrine–adfluvial or allacustrine and anadromous, but the latter vary among catchments as to which is favoured. Indirect evidence suggests that around 50% of the variability in S. trutta migration v. residency, among individuals within a population, is due to genetic variance. This dichotomous decision can best be explained by the threshold‐trait model of quantitative genetics. Thus, an individual's physiological condition (e.g., energy status) as regulated by environmental factors, genes and non‐genetic parental effects, acts as the cue. The magnitude of this cue relative to a genetically predetermined individual threshold, governs whether it will migrate or sexually mature as a river‐resident. This decision threshold occurs early in life and, if the choice is to migrate, a second threshold probably follows determining the age and timing of migration. Migration destination (mainstem river, lake, or sea) also appears to be genetically programmed. Decisions to migrate and ultimate destination result in a number of subsequent consequential changes such as parr–smolt transformation, sexual maturity and return migration. Strong associations with one or a few genes have been found for most aspects of the migratory syndrome and indirect evidence supports genetic involvement in all parts. Thus, migratory and resident life histories potentially evolve as a result of natural and anthropogenic environmental changes, which alter relative survival and reproduction. Knowledge of genetic determinants of the various components of migration in S. trutta lags substantially behind that of Oncorhynchus mykiss and other salmonines. Identification of genetic markers linked to migration components and especially to the migration–residency decision, is a prerequisite for facilitating detailed empirical studies. In order to predict effectively, through modelling, the effects of environmental changes, quantification of the relative fitness of different migratory traits and of their heritabilities, across a range of environmental conditions, is also urgently required in the face of the increasing pace of such changes.

balance between enhanced feeding and thus growth advantages of migration to a particular habitat v. the costs of potentially greater mortality and energy expenditure.
Fluvial-adfluvial migration usually has less feeding improvement, but less mortality risk, than lacustrine-adfluvial or allacustrine and anadromous, but the latter vary among catchments as to which is favoured. Indirect evidence suggests that around 50% of the variability in S. trutta migration v. residency, among individuals within a population, is due to genetic variance. This dichotomous decision can best be explained by the threshold-trait model of quantitative genetics. Thus, an individual's physiological condition (e.g., energy status) as regulated by environmental factors, genes and non-genetic parental effects, acts as the cue. The magnitude of this cue relative to a genetically predetermined individual threshold, governs whether it will migrate or sexually mature as a river-resident. This decision threshold occurs early in life and, if the choice is to migrate, a second threshold probably follows determining the age and timing of migration. Migration destination (mainstem river, lake, or sea) also appears to be genetically programmed. Decisions to migrate and ultimate destination result in a number of subsequent consequential changes such as parr-smolt transformation, sexual maturity and return migration. Strong associations with one or a few genes have been found for most aspects of the migratory syndrome and indirect evidence supports genetic involvement in all parts. Thus, migratory and resident life histories potentially evolve as a result of natural and anthropogenic environmental changes, which alter relative survival and reproduction. Knowledge of genetic determinants of the various components of migration in S. trutta lags substantially behind that of Oncorhynchus mykiss and other salmonines. Identification of genetic markers linked to migration components and especially to the migration-residency

| INTRODUCTION
Migration occurs in all major animal taxa and results from spatial, seasonal and ontogenetic separation of optimal habitats for feeding and breeding (Northcote, 1984). However, the spatial patterns and behaviours involved vary enormously among species, populations and individuals (Dingle & Drake, 2007). Better understanding of migration requires studies of convergent processes across a wide range of taxa (Dingle, 2014;Sahashi & Morita, 2013). In broad terms, the study of migratory syndromes, the integrated suites of traits, behaviours and physiological processes involved directly or indirectly in migration (Dingle, 2006;van Noordwijk et al., 2006), can be approached from both proximate and ultimate perspectives (Tinbergen, 1963). Proximate questions concern how migratory tendencies, behaviours or associated traits are expressed in individuals in response to environmental cues or constraints during ontogeny. Ultimate questions focus instead on the evolutionary functions and phylogenetic history of migration. The proximate mechanisms themselves, however, have evolved in response to past environmental pressures and can evolve further as selective regimes change. In recent years there has been an increasing realisation that genetic mechanisms play a major role in the control of migratory behaviour in a wide range of animals and that a study of this genetic architecture enhances our understanding of the mechanisms involved (Liedvogel et al., 2011). It is also essential to understand how natural selection operates at various levels in the complex chain linking genes to phenotypes to Darwinian fitness in variable environments. These insights can then feed into a more evolutionarilyenlightened approach to the conservation and management of migratory species, which face multiple anthropogenic threats worldwide.
While some species have obligate migratory or non-migratory life histories, others exhibit intraspecific variation in migratory tendencies, with populations in some parts of the range being fully migratory, others being fully resident and yet others exhibiting a facultative mix of migratory and resident individuals ('partial migration' of some authors; Chapman et al., 2011). Fishes provide many interesting examples here, both in terms of population and individual-level variation in migratory tendencies, but also in the habitats and environments to which fish migrate (Chapman et al., 2012). Salmonids are particularly interesting in this regard as they can exhibit large or short distance migrations or fully resident life histories (Dodson et al., 2013). Migrations can be between fresh water and salt water or confined to lakes and rivers. Like any complex phenotype, variais possible when precise destination and life-history details, including age and timing of various events and repeat spawning, are taken into account . Salmo trutta can be resident within rivers, often in a 1st or 2nd order tributary, for their entire life cycle; i.e., river-resident. Included within this term are individuals that make early localised dispersal movements, as described for example by Vøllestad et al. (2012). Salmo trutta can also be resident within lakes with their entire life cycle being spent there; i.e., lake-resident. This life history is probably more common than hitherto recognised. However, there may be both horizontal and vertical movements within lakes between spawning and feeding grounds and on a diurnal basis (Jonsson & Jonsson, 2018). Lake-spawned S. trutta appear to remain resident and do not migrate to the river or sea, as, for example, occurs with some sockeye salmon Oncorhynchus nerka (Walbaum 1792); although this aspect has not been specifically investigated. Some authors use the term freshwater-resident in the sense of riverresident only, while others use it in the sense of inhabiting to include all freshwater forms including migratory ones. Because of this ambiguity, the term should not be used but, instead, precise life history should be specified (Ferguson et al., 2017).
In this review migration refers to directed movements between two distinct habitats occurring with regular periodicity on a temporally predictable basis (Brönmark et al., 2014;Northcote, 1978). Outmigration typically takes place for feeding, or to find temporary refuge. Thus, extreme temperature and water flow in the natal river can result in S. trutta moving downstream to find refuge, presumably as a direct result of stress. Out-migration is followed by a return migration to the place of natal origin for subsequent spawning, or to the natal or non-natal area for refuge from harsh conditions (Klemetsen et al., 2003;Jensen et al., 2015). The terminology used here for potamodromous (freshwater) migrations follows that of Varley and Gresswell (1988) as elaborated by Northcote (1997). Salmo trutta resulting from spawning in rivers can undergo three main types of migration to feeding areas ( Figure 1) and subsequent adult return, involving: (a) a larger tributary or typically the main stem of the river, fluvial-adfluvial migration; (b) a lake, lacustrine-adfluvial migration if an inlet river is involved, or allacustrine migration where outlet river spawning occurs; and (c) the sea, anadromy. In the latter case movement may only be as far as the estuary, with some authors referring to this as semi-anadromy or partial-anadromy, but, confusingly, these terms are also used by other authors in the sense of facultative migration. In Denmark and Norway migration can be terminated in a fjord rather than continue to the open sea (del Villar-Guerra et al., 2014; Thorbjørnsen et al., 2018). Some authors (Quinn, 2018) use a more abbreviated terminology for potamodromy simply referring to adfluvial for migrations between natal rivers and lakes and fluvial both for migrations within rivers and river-residents.
In the literature there are many references to migratory S. trutta in the sense of anadromy only. However, potamodromous migrations are widespread and in many parts of the range such migrations are more numerous or are the only migrations present. A fundamental thesis of this review is that fluvial-adfluvial, lacustrine-adfluvial, allacustrine and anadromous migrations are fundamentally the same as to their determinants and thus information on one form of migration is relevant to the others. It is important, however, in this context to separate the decision to migrate from the decision as to the destination of migration. The exclusive focus on anadromy in many studies is probably due to the commercial and recreational importance of anadromous O. mykiss in western North America and anadromous S. trutta in north-western Europe, where most studies have been undertaken, rather than any major difference in their migration. Comparatively few studies have been carried out into the determinants of lacustrine-adfluvial and allacustrine migrations in salmonines and even less on fluvial-adfluvial migration. This bias is inevitably reflected in the relative coverage here. Here the term migrant is used where all types of migration are being referred to but otherwise qualified. Due to the considerable similarity in determinants and processes outlined below and in Table 1, the term smolt is not restricted to destinations involving hypo-osmoregulation and is used here for all downstream migrating juveniles irrespective of their ultimate destination, as has also been applied by other authors Jones et al., 2015) and indeed has been widely used for migratory S. trutta in the Baltic Sea.

| Life-history occurrence and patterns
In the formerly glaciated region of north-western Europe, as a result of marine barriers after the ice retreated, most current freshwater S. trutta populations are derived from anadromous ancestors Lake Sea F I G U R E 1 Potential life-history diversity of Salmo trutta in a typical catchment with a lake. , Spawning locations; , adult feeding sites. ( ) Lake-or river-resident, ( ) Fluvial-adfluvial, ( ) Lacustrine-adfluvial, ( ) Allacustrine, ( ) Semi-anadromous, and ( ) Anadromous (Ferguson, 2006 Spawning of lake-resident S. trutta occurs on shoreline gravels where there is sufficient wave action or diffuse water flow from the surrounding land to provide oxygenation. This type of spawning is typical of many small upland lakes (Prodöhl et al., in press). Spawning has been shown to occur in gravel areas in large lowland Irish lakes such as Lough Melvin (54 24 0 N, 08 07 0 W; Ferguson & Taggart, 1991) and Lough Mask (53 36 0 N, 09 22 0 W; P. Gargan, Inland Fisheries Ireland, pers. comm.). Spawning can also occur deep within lakes where there is upwelling from bottom fissures. The latter is typical of spawning in many volcanic-region lakes in Iceland where water flow is underground  and in lakes in limestone areas elsewhere. In Lake Garda (Italy; 45 34 0 N, 10 38 0 E), native S. trutta spawning has been shown to occur at 200-300 m depth and in Lake Posta Fibreno (Italy; 41 41 0 N, 13 41 0 E) spawning occurs in underground spring-fed karstic pools (Meraner & Gandolfi, 2018).
Fluvial-adfluvial, lacustrine-adfluvial and allacustrine migratory S. trutta occur throughout the native range where suitable conditions exist. Genetic assignment studies have indicated that S. trutta feeding in the mainstems of some Irish rivers are entirely composed of recruits from tributaries; i.e., they are fluvial-adfluvial migrants . In regions such as Ireland and Scotland, UK, with thousands of freshwater lakes, a lacustrine-adfluvial life history is numerically the most common one, based on the relative abundance of such populations . Allacustrine populations are also widespread and are typically reproductively isolated and genetically distinct from the lacustrine-adfluvial populations of the same lake (Ferguson, 2004 (Dębowski, 2018;Kallio-Nyberg et al., 2010;Soininen et al., 2018). Anadromous S. trutta are found in the Black and Caspian Sea drainages (Makhrov et al., 2018) and formerly also in the Aral Sea prior to desiccation and salinity increase (Markevich & Esin, 2019). Anadromous S. trutta are currently absent from Mediterranean rivers, most likely because of the high temperature of the sea, although the widespread distribution of S. trutta in many unconnected catchments indicates that anadromy occurred during glacial periods when the sea temperature was lower (Gibertoni et al., 2014).
Major physiological differences between potamodromy and anadromy are the changes required for osmoregulation and ionic regulation. Thus, in fresh water the body fluids of S. trutta and other teleosts are hyperosmotic and hypotonic to the surrounding water and are faced with the gain of water by osmosis and the loss of ions by diffusion, with the reverse being the case in full-strength seawater.
Anadromy thus requires a change in regulation when moving between fresh and seawater to maintain osmotic and ionic homeostasis. This is achieved through osmosensing, which is the physiological process of perceiving a change in environmental salinity and with which many genes have been found to be associated (Kültz, 2013). Teleosts maintain their internal salt concentration at around one-quarter to one-  (Klemetsen et al., 2003), a phenomenon also known in O. mykiss (Null et al., 2013) and dolly varden charr Salvelinus malma (Walbaum 1792), where older individuals cease to migrate (Bond et al., 2015). It is increasingly recognised that anadromous S. trutta can spend a lesser part of their life at sea with the rest spent in lakes or rivers. In a Norwegian tracking study involving previously spawned migrants (kelts), variation in marine residence ranged from 7 to 183 days, this residence being positively correlated with size and original smolt age and negatively with date of sea entry (Eldøy et al., 2015). In Loch Lomond (Scotland; 56 05 0 N, 04 36 0 W), a 71 km 2 freshwater lake, carbon stable-isotope analysis showed that individual S. trutta appear to move repeatedly between the lake and estuarinemarine environments (Etheridge et al., 2008). In this case there is only a short river (c. 10 km) separating the loch and the Clyde Estuary and it may be that migration does not go beyond the estuary. In the

| WHY MIGRATE AND WHERE?
Salmo trutta populations in many rivers show facultative migration, with part of the population migrating while other individuals remain resident within their natal river. Migration potentially offers many benefits to individuals while at the same time these are countered by ensuing costs (Brönmark et al., 2014;Gross, 1987;Quinn & Myers, 2004), resulting in the outcome being finely balanced between these conflicting aspects (Ferguson et al., 2017). Advantages and disadvantages are likely to vary among populations and temporally as a result of environmental and biotic changes; e.g., population density. The increasing food availability hypothesis  explains why salmonines migrate from natal areas, with a balance of relative risks and rewards determining where they migrate to. Migration downstream within a river system, to a lake, or to the sea probably increases feeding opportunities. Mechanisms driving these migrations are probably the same as long as productivity between natal river and feeding habitats is significantly different (Ayer et al., 2017). Better feeding, both in terms of quantity and quality, results in faster growth, potentially larger size at maturity, higher fecundity, greater energy stores at reproduction and thus more offspring are produced (Acolas et al., 2008;Fleming & Reynolds, 2004;Jonsson & Jonsson, 2006a). Goodwin et al. (2016) found that the parental contribution of males and, especially, females to the juvenile production in a river was much higher for anadromous than river-resident S. trutta.
Based on size at maturity, lake and sea feeding is superior to remaining within the river, albeit the relative importance of lake and sea feeding varies among catchments. If lacustrine S. trutta become piscivorous (Campbell, 1979;Wollebaek et al., 2018), they can reach a larger size than anadromous conspecifics. Thus, in both Britain and Ireland, the largest rod-caught piscivorous lacustrine S. trutta to date have had a greater mass than the largest anadromous S. trutta (Ferguson et al., 2017), although the abundance of prey fish is such that only a small proportion of individuals can adopt piscivory (Campbell, 1979;Hughes et al., 2016)  . The largest S. trutta known were the socalled salmon of the Caspian Sea with the largest recorded being 57 kg (Markevich & Esin, 2019), although their size has been decreasing in recent decades (Niksirat & Abdoli, 2009). Large S. trutta are also known from the Baltic Sea (Rasmussen & Pedersen, 2018). Possibly, the brackish nature of these seas results in less energy expenditure for osmoregulation than in either fresh water or full-strength seawater (see §1.1).
On the adverse side, migration increases energy expenditure, physiological stress, risk of predation, parasites and diseases, both during migration and in the subsequent habitat. Migration downstream within a river system, to a lake, or to the sea increases risk in that order. The number of S. trutta predators appears to be higher at sea than in fresh water ) and predation is a major mortality factor shortly after smolts reach the sea (Dieperink et al., 2002;Healy et al., 2017). Predation in lakes is generally higher than that in rivers (Schwinn et al., 2018) and especially at river-lake confluences (Kennedy et al., 2018).
Lacustrine-adfluvial migration is probably similar or better in benefits to anadromy in some cases but lowers the relative costs due to lowered energy expenditure and decreased risk of predation. Anadromous S. trutta occur especially in shorter river systems of low alkalinity with good spawning and nursery areas easily accessible from the sea and especially where river or lake productivity is low (CSTP, 2016). In higher productivity lakes, lacustrine-adfluvial migration can occur exclusively even where there is no barrier to anadromy suggesting that, in that situation, it is superior in terms of cost-benefit considerations. In other cases, both lacustrine-adfluvial and anadromous S. trutta are present in the same catchment, with the lacustrine-adfluvial form often predominating (Poole et al., 2006). In O. mykiss, anadromy is also less common in river systems with large lakes (Kendall et al., 2015). In some situations, S. trutta migration may occur to downstream brackish lakes or estuaries. Thus, in the Burrishoole system (western Ireland) many smolts migrate to the brackish Lough Furnace (53 55 0 N, 09 35 0 W) and appear to remain there, or in the estuary, before returning to fresh water (Poole et al., 2006). It has been suggested that estuaries provide better feeding than rivers but with reduced likelihood of predation and reduced salinity compared with the open sea (Thorpe, 1994), although fluctuating salinity may actually produce greater physiological stress than the higher, but more stable, salinity of seawater .

| Sex and facultative migration
Many studies have shown that in S. trutta populations there is generally a sex bias, with typically females predominating among fluvialadfluvial, lacustrine-adfluvial and anadromous migrants and males among residents (Ayer et al., 2017;Ferguson et al., 2017;García-Vega et al., 2018;Huusko et al., 2018). A sex bias is to be expected from the balance of benefits of migration and residency (Hendry et al., 2004). Thus, female reproductive success is generally limited by gamete production with a larger body size giving greater fecundity and egg size (Fleming, 1996;Quinn, 2018). Larger females can attract mates, acquire and defend better spawning sites in a wider range of substrate sizes and excavate deeper nests (Fleming & Reynolds, 2004).
Compared with females, male reproductive success is typically limited by access to mates (Fleming, 1998) rather than gamete production, since even small males can produce millions of sperm (Munkittrick & Moccia, 1987). While a larger size can be of benefit to males in attracting and defending mates, obtaining a large body size is less critical for male reproduction because instead of aggressive defence of females, a tactic typically displayed by larger anadromous males (Esteve, 2005), they can adopt a sneaking tactic allowing successful egg fertilisation at a small size (Gross, 1985). Thus, males more often mature as residents since they are less dependent on large body size for reproductive success and, consequently, mature across a much greater range of ages and sizes (Jonsson & Jonsson, 1993). Early maturity in males also results in reduced pre-reproductive mortality (Gross & Repka, 1988). In Atlantic salmon Salmo salar L. 1758, male parr may mature while still in fresh water and then subsequently undergo an anadromous life history (Mitans, 1973). In S. trutta, maturation appears to exclude subsequent anadromy (Dębowski & Dobosz, 2016). However, L'Abée-Lund et al. measured sex-bias in gene expression in the brain transcriptome of O. mykiss in two F 1 lines derived from migratory and resident fish, which were reared in a common-garden environment to reveal heritable differences. The parents came from Sashin Creek (Alaska; 56 21 0 N, 134 43 0 W), the residents being from an above-waterfalls population that had been artificially established from the anadromous stock below the waterfalls some 70 years previously (Thrower et al., 2004). Overall 1716 genes (4.6% of total examined) showed evidence of sex-biased gene expression involving at least one time point from the fry stage through to when they either migrated to the ocean or remained resident and became sexually mature. The majority (96.7%) of sex-biased genes were differentially expressed during the second year of development, indicating that patterns of sex-bias in expression are linked to key developmental events, such as migration and sexual maturation. This is not surprising as the brain is involved in hormonal regulation of both of these processes (Hale et al., 2018).

| Reproductive isolation and heritability
The key behavioural step for a young S. trutta in its natal river is whether to remain in the river and become sexually mature or migrate to a higher-order tributary, a lake, or the sea. While offspring of migratory and river-resident S. trutta can show different life histories from their parents, there is often a strong tendency to track the parental life history (Dębowski & Dobosz, 2016;Jonsson, 1982;Skrochowska 1969). Berejikian et al. (2014) showed that female offspring produced by anadromous O. mykiss mothers rarely expressed residency (2%), while the percentage of maturing male parr produced was much higher (41%) across a diversity of freshwater habitats. Also, both male and female parr that were produced by resident mothers were significantly more likely to show residency than the offspring of anadromous mothers. Female body size has a significant effect on egg size, a heritable trait (Carlson & Seamons, 2008), which affects survival and growth of juveniles, especially in the early stages of life (Thorn & Morbey, 2018). Associations between maternal and offspring life-histories could therefore reflect a mix of direct genetic effects (e.g., where offspring inherit migratory alleles from their mother), indirect genetic effects (e.g., where genes carried by the mother affect the size of her eggs, which in turn influences offspring life history) and maternal environmental effects (where environmental effects on the mother's phenotype influence offspring life history).
Most studies with neutral genetic markers have failed to find genetic differentiation between strictly sympatric (syntopic) migratory and river-resident S. trutta (Ferguson et al., 2017). However, this does not imply that the difference between these life histories does not have a genetic basis, only that it is not population based. Thus, most genetically-based characteristics are inherited through family lineages.
There are many cases of neutral genetic markers showing genetically distinct resident S. trutta above complete or partial barriers that are impassable or have restricted passage to upstream migration, compared with migratory ones below, as would be expected from allopatric populations where gene flow is limited or absent (Ferguson, 1989).
In addition, there are also situations of upstream resident and downstream facultative migratory S. trutta populations within the same river but without any physical barrier to upstream movement (Lemopoulos et al., 2017), with the two types representing separate colonising lineages in some cases (Hamilton et al., 1989;McKeown et al., 2010;Turan et al., 2009). In some situations of sympatric differentiation below barriers, the resident S. trutta or O. mykiss appear to have arisen from displacement of such fish from above a barrier where strong selection for residency is expected (Ferguson et al., 2017;Pearse et al., 2009). Where migratory and river-resident salmonines occur in syntopy, behavioural or temporal differences in spawning could result in reproductive isolation.
Overall the evidence suggests that there is a strong genetic ele- traits (Dodson et al., 2013;Roff, 1996), it is likely that around half of the phenotypic variability in S. trutta migration v. residency among individuals within a population is due to additive genetic variance, with the remainder attributed to non-additive genetic variance, nongenetic parental effects and environmental influences. However, it is very important to acknowledge that heritability estimates are specific to the population and particular environmental conditions examined.
Explicit estimates are required for a range of S. trutta populations under different conditions before credence is given to any estimate.
The same also applies to genetic correlations among traits and it may be the case that patterns of phenotypic integration (when multiple functionally-related traits are correlated with each other, in part due to pleiotropic effects of genes) may be quite different in S. trutta compared with O. mykiss, with some aspects being less restricted to evolve independently than others. of selection related to lacustrine-adfluvial migratory v. river-residency in S. trutta in two catchments in Finland. Interestingly, four of the eight outlier SNPs mapped to genes previously shown to be involved in anadromy in salmonines with the three others being associated with genes involved in temperature changes and feeding.

| Physiological condition
Numerous salmonid studies have linked many facets of an individual's physiological condition with the decision to migrate or remain resident, including aspects such as energetic state, metabolic rate and lipid storage (adiposity) levels (Ferguson et al., 2017). food quality (i.e. energy value) may be as important as food quantity (Kendall et al., 2015). Given the relation with food availability, it might be expected that an increase in juvenile density would lead to a greater propensity for migration if competition for resources is present. Montorio et al. (2018) found that for S. trutta, first-year density showed no correlation with migration although it correlated negatively with first-winter survival and body size; the latter potentially resulting in delayed migration (see below). However, S. salar density was found to be positively correlated with S. trutta migration, indicating interspecific competition.
Temperature appears to be a key abiotic factor in the migratory decision (Brannon et al., 2004;Sloat & Reeves, 2014), with both absolute temperature and variation in temperature being important (Kendall et al., 2015). Temperature is clearly linked to food availability,   (Roff, 1996). This has been variously described in

| Does size matter?
In field experiments, various aspects of size such as length, growth rate, body mass and condition-factor (length-mass ratio) are often used as surrogates for physiological condition, albeit the evidence for this association being inconclusive. Use of size as a surrogate potentially confuses two apparently separate thresholds. Thus, in S. trutta there appears to be an early threshold related to the decision to migrate and a second threshold linked to the actual timing (i.e., age) of migration, the latter remaining flexible for a longer period and potentially affected by environmental changes subsequent to the decision to migrate . That is, individuals that have taken the decision to migrate may then have to pass a population-specific size threshold before migration occurs, if not migration is deferred, resulting in migration occurring at different ages ( Figure 2). Survival of migrants on entry to the new environment (e.g., lake or sea) may be positively size-dependent in both S. trutta and O. mykiss (Klemetsen et al., 2003;Phillis et al., 2016), while larger fish that defer their migration may meanwhile fail to meet their higher energetic requirements in the river. Thus, larger fish at this second decision window should be selected to migrate now and smaller fish to defer, with the actual size threshold that evolves in a given population depending on the local selective pressures. In contrast, there is no reason for the initial migration v. residency decision, which occurs earlier in ontogeny, to involve a size threshold. Juveniles of a given genotype that encounter poor early feeding conditions are expected to be more likely to choose migration, but they could actually be smaller at this first decision window than other individuals that encountered better early feeding. Or there may be no obvious size difference between fish choosing migration v. residency at this point, despite differences in physiological condition. Confusingly, some studies appear to have been looking at the size threshold for timing of migration rather than the threshold for the migrate-mature decision (e.g. Phillis et al., 2016).
Not surprisingly then, size has been linked to the propensity for migration both positively and negatively and with evidence of population specific responses (Jonsson, 1985). Compared with residents, migratory S. trutta individuals have been found to be smaller (Morinville & Rasmussen, 2003) or larger (Acolas et al., 2012), have lower body mass (Winter et al., 2016), have lower condition-factors (Boel et al., 2014;Wysujack et al., 2009)  Conflicting results may also in part be due to failure in some studies to fully account for potentially confounding variables. For example, size at migration is unlikely to reflect size at decision time perhaps a year earlier (Acolas et al., 2012;Beakes et al., 2010;McKinney et al., 2015). In the meantime residents may have diverted energy from growth to sexual maturation. As survival in the early marine phase is size dependent (Klemetsen et al., 2003;Phillis et al., 2016), premigrants may have accelerated growth during this period with the extent of growth being negatively correlated with size at last annulus (Thomson & Lyndon, 2018). That is, potentially the largest smolts may

| MIGRATION DESTINATION
Once the decisions are taken regarding if and when to migrate the next decision is, the destination for adult feeding. As indicated by the results described below, while more attention has been given to the mechanisms of return spawning migration and natal homing (Bett & Hinch, 2016), various indirect lines of evidence suggest that outmigration pathways in S. trutta and other salmonines are also genetically influenced. In passerine birds migration pathways to geographically distinct wintering areas are genetically encoded and specific genes associated with particular migratory phenotypes have been identified for some species (Lundberg et al., 2017). However, while salmonines show innate compass orientation in the marine phase (see below), it is not known if the resolution of the magnetic-field map is sufficient to provide positional information over the more limited scale of a river catchment (Scanlan et al., 2018). In some situations, it is not a matter of moving downstream until the feeding destination is reached since, for some allacustrine populations where spawning occurs in a tributary of the outlet, getting to the lake requires downstream migration followed by upstream migration (Figure 1). It is difficult to envisage how this could be achieved without innate instructions. Allacustrine spawning salmonines must move upstream to reach the lake unlike downstream migrating lacustrine-adfluvial inlet spawners. Several common-garden experimental studies on S. trutta, O. mykiss and cutthroat trout Oncorhynchus clarkii (Richardson 1837) have indicated that this, as with the downstream movement of inlet-spawned offspring, is an inherited adaptive response to current direction (Jonsson et al., 1994;Kelso et al., 1981;Raleigh & Chapman, 1971).
In general, for a particular S. trutta population the feeding destination remains fixed from year to year although it can change over time as a result of natural selection due to alterations in costs v. benefits for migration to that particular habitat (see §7) . The destination appears to be already decided when the migration begins. Anadromous S. trutta can move through both lake and downstream river habitats to reach the sea without any indication of stopping on route.
Similarly, lacustrine-adfluvial S. trutta can migrate through other lakes to reach their destination lake . Given that both mortality and energy expenditure of salmonid smolts are considerably distance that individual S. trutta migrate may be controlled by energy status. Thus, short-distance lacustrine-adfluvial migrants were more lipid depleted than long-distance, potentially anadromous, migrants that continued their migration through the lake. The fish with greater energy depletion apparently terminated their migration at the earliest increased feeding opportunity. These studies would suggest that environmental factors, such as food availability in relation to metabolic needs, play a part in determining migration destination. However, along with environmental factors, genes play a role in metabolic efficiency and energy status and thus may indirectly determine migration destination in these two catchments.
There is good evidence for genetic control of feeding location in the marine phase for anadromous salmonines with several species, including S. trutta, showing site fidelity for feeding location (Losee O. tshawytscha were shown to use an inherited magnetic map that facilitates navigation during their oceanic migration (Putman et al., 2014b). Salmo salar, from a long-standing non-anadromous population, were shown to be able to orientate in novel magnetic fields (Scanlan et al., 2018). As this ability to extract location information from the Earth's magnetic field is present in at least three salmonines species, it seems to be an ancestral state in the sub-family and thus is very likely to be present in S. trutta.

| CAUSES AND CONSEQUENCES OF LIFE-HISTORY DECISIONS
Fundamental to facultative migration is the decision on whether to migrate or to remain as a resident in the river and mature, which may take place a considerable time before external evidence of migration occurs (Hecht et al., 2015;McKinney et al., 2015). The switch between resident and migratory phenotypes is a complicated process involving sensing the cue, comparing it to an individual threshold, triggering a physiological or other response and development of that response (Buzatto et al., 2015). It is important to distinguish between the decision-making process and the many subsequent responses activated by that decision. Failure to recognise that important distinction has led to misinterpretation of some studies. Studies at the smoltification stage (see §6.2) are the earliest at which it is possible to externally differentiate migrants from residents within a population and many comparative studies on smolts and non-smolts have been undertaken for this practical reason. However, such studies primarily indicate the physiological and other changes necessary for migration or maturation and not with why the decision to migrate was taken in the first place. Similarly, it is important to distinguish between environmental factors involved in the migration decision from those that act as stimuli for the timing of the actual migration. Pirhonen et al. (1998) found that both anadromous and lacustrine-adfluvial S. trutta smolts migrated at the same time, suggesting that similar influences may be involved in their timing.

| Gene regulation and epigenetics
The translation of the same genome into different phenotypes (phenotypic plasticity) requires differential gene expression. That is, the genotype does not unambiguously determine the phenotype but rather the range of phenotypes that can be produced under different environmental conditions. This is referred to as the reaction norm.
Gene regulation involves various chemical messages that are responsible for switching individual genes on or off, thus facilitating or inhibiting the production of specific proteins, but without changing the underlying DNA sequence. Collectively these changes are referred to as epigenetic mechanisms, with the modified genome being referred to as the epigenome. However, the term epigenetic is often used inconsistently and is sometimes used synonymously with epigenetic inheritance, which is a separate process ( A key but underappreciated aspect is that theory generally predicts (Bonduriansky & Day, 2009) that transgenerational non-genetic effects, including the special case of epigenetic inheritance, should only be adaptive when there is some degree of predictability or autocorrelation between the parental and offspring (or grand-offspring, etc.) environments. If the environment in generation t + 1 is uncorrelated or only weakly correlated with the environment in generation t, then trans-generational inheritance of environmental effects will be of little adaptive value and could actually increase the likelihood of phenotype-environment mismatching (Burton & Metcalfe, 2014). Similar arguments apply to within-generation phenotypic plasticity, which is only adaptive when the environment at the time of Creek reared under communal hatchery conditions for 1 year. They found differential gene expression in the brain between these lines for 1982 genes (7% of genes examined). Differences between anadromous and resident offspring were detected from hatching onwards with the greatest number of gene differences being found at 8 months of age, more than a year before obvious external appearance of smolt transformation. Patterns of gene expression during development differed between males and females, which may reflect the fact that males, in the resident population, mature earlier than females  Giger et al. (2008) found that 21% of screened genes were differentially expressed in S. trutta smolts and non-smolts, which would suggest that many genes are involved in smoltification, or are indirectly affected by the process, in keeping with the genome-wide distribution of gene associations found in later studies. Many other studies have shown gene expression differences, especially in the gills, between smolts and resident salmonines (Houde et al., 2018;McKinney et al., 2015;Veale and Russello, 2017a). Baerwald et al. (2015) found 57 DMRs between smolt and resident O. mykiss juveniles derived from a cross reared under communal conditions. Genes that have been found to be differentially expressed relate, in most cases, to known physiological differences between smolts and residents; i.e., those associated with circadian rhythms, growth, homing, innate immunity, light sensitivity, metabolism, morphology, olfactory imprinting, osmoregulation and sexual maturation. Transaldolase 1 and endozopine are expressed at lower levels in both potamodromous and anadromous individuals compared with resident individuals and these differences can be detected some 3 months prior to migration (Amstutz et al., 2006;Giger et al., 2008); emphasising again the commonality of migration irrespective of destination.

| Smoltification
Smoltification is a universal feature of all migratory salmonines and involves many changes including alterations to salmonid body shape and behaviour, silvering and changes to many enzymes and hormones, especially those produced by the thyroid (McCormick, 2013).
Although often seen only as a preparation for anadromy, there is increasing evidence that many similar changes occur for potamodromous migrants. Thus, increase in NKA activity, widely used as an indicator of anadromous smoltification, occurs also in potamodromous migrants (Boel et al., 2014;Inatani et al., 2018), as do changes in behaviour, skin pigmentation and body morphology (Table 1). Whether this increase in NKA in potamodromous migrants has functional significance or simply reflects ancestral standing genetic variation is not known. NKA is composed of two structural subunits, α and β, together with a regulatory subunit, γ (Blanco & Mercer, 1998). In salmonines there are five isoforms of the NKA α subunit with α1a producing an NKA isozyme suited to fresh water and ion uptake and α1b suited to marine conditions and ion excretion (McCormick et al., 2009). Thus, in anadromous salmonines at parr-tosmolt transformation there is a switch in the α subunit composition, in addition to an overall increase in NKA activity, . Downregulation of the α1a subunit and upregulation of the α1b occur while the fish are still in fresh water and this occurs prior to the increase in NKA activity in S. trutta (Seidelin et al., 2000), indicating that the migration destination is pre-determined. Nonanadromous Oncorhynchus masou (Brevoort 1856) were found to show an increase in NKA activity in smolt-like individuals but, unlike anadromous individuals, this was not accompanied by an increase in the α1b isoform (Inatani et al., 2018).
A recent laboratory study of S. trutta showed that offspring of wild-caught parents deriving from a naturally non-anadromous population in Western Ireland displayed morphological signs of smoltification when exposed to reduced food supply as fry/parr, compared with fish from the same population experiencing optimal food rations (Archer et al., in press). However, putative smolts from this non-anadromous population background exhibited reduced saltwater tolerance (as assessed by plasma chloride levels following 24 h of saltwater exposure) compared with smolts from a second population, which exhibits high rates of anadromy in the wild despite both sets of smolts having being raised under identical experimental conditions. These findings indicate that non-anadromous wild populations of S. trutta may retain some genetic capacity for facultative anadromy, albeit with imperfect saltwater tolerance among resulting smolts, as has also been shown in O. mykiss (Phillis et al., 2016;Thrower et al., 2004).
Quantitative trait locus (QTL) studies show that variation in salinity tolerance among individuals of S. salar, O. mykiss and Arctic charr Salvelinus alpinus (L. 1758) has a genetic basis, with the same genes being involved in these species (Norman et al., 2012). The timing of smoltification is in response to environmental cues such as photoperiod (Strand et al., 2018), temperature (Haraldstad et al., 2017) and water flow , with the brain being the main integrator of this information and thus, the main regulator of the process emphasising that these are two distinct processes.

| Return migration
Out-migration requires a subsequent in-migration for spawning or, where adverse conditions occur at the migration destination, to a suitable refuge from harsh conditions. Non-mature fluvial-adfluvial S. trutta in Spain have been shown to migrate upstream at times other than the main spawning run period, possibly for thermoregulation (García-Vega et al., 2018). In Norway, where low winter sea temperatures occur, overwintering of immature anadromous S. trutta is often in fresh water (Klemetsen et al., 2003). This is possibly due to the marine hypo-osmoregulatory capacity being compromised by low temperature, although, it may also reflect differing life-history traits or individual genetic differences in osmoregulatory capacity, since not necessarily all S. trutta in a population exhibit the behaviour (Thomsen et al., 2007) or individual genetic differences in osmoregulatory capacity. Thus, there are population differences in the expression of key stress and osmoregulatory genes suggesting that some populations may be more adapted to remaining at sea overwinter than others (Larsen et al., 2008) and there may also be individual heritable differences within populations. Where the return is for overwintering without spawning, both natal and non-natal rivers are used. Anadromous S. trutta in Norway have been recorded wintering up to four times in other rivers before returning to their natal one for reproduction (Jensen et al., 2015). Studies on S. alpinus indicate that they overwinter in the closest rivers with the least energetically demanding migratory route, thereby potentially minimizing the migration costs in nonbreeding years (Moore et al., 2017). Overwintering in non-natal rivers means that individual movements of physically tagged S. trutta overestimate the extent of actual gene flow among populations in different catchments (Masson et al., 2018) and that samples of older post-smolt individuals are inappropriate as baseline samples in genetic assignment studies (Moore et al., 2017).
Return for spawning is generally to the natal river with this homing being undertaken with considerable accuracy as shown by the typical population genetic structuring of both anadromous and potamodromous S. trutta (Ferguson, 1989;Prodöhl et al., 2017). Longer distance homing at sea likely involves geomagnetic fields but closer to the home catchment and within the catchment, olfactory cues derived from the chemical composition of the natal river or population-specific pheromones are important (Bett & Hinch, 2016).
Social interactions; i.e., migrating as a group, may also play a part in navigation (Berdahl et al., 2014).
The age at which maturation and spawning migrations occur, together with the time of year and ultimate location of spawning vary among individuals within populations and among populations of salmonines, including S. trutta (Klemetsen et al., 2003;Quinn et al., 2016). Variation in all of these aspects has been shown to have significant genetic components consistent with evidence of local adaptation in these life-history traits. Thus, age of maturation in S. trutta and other salmonines is a quantitative trait (Palm & Ryman, 1999) with a moderate heritability (Dodson et al., 2013;Reed et al., 2018). Variation in the incidence of maturation of male S. salar as parr has also been shown to be substantially genetically controlled (Lepais et al., 2017). In S. salar, a single locus containing the vgll3 gene, with sexdependent dominance, has been shown to explain 39% of the differences in sea age at maturity (Ayllon et al., 2015;Barson et al., 2015).
This locus has also been linked with iteroparity, with the early maturing genotype being more likely to reproduce again (Aykanat et al., 2019). Interestingly, the vgll3 locus has previously been found to be associated with the timing of puberty in humans, suggesting a conserved mechanism for timing of maturation in vertebrates (Kjaerner-Semb et al., 2018). It would therefore seem highly likely that the same gene region may control time of maturity in S. trutta. Indeed, the two non-synonymous substitutions identified in vgll3 in S. salar are also present in S. trutta (Ayllon et al., 2015).
Although return migration timing within the year can be closely associated with spawning time, this is not always the case. Return migration timing has been shown to have a heritable basis in many salmonid species (Cauwelier et al., 2017) with clock genes being a signifi- However, in a denser gene mapping study, Micheletti et al. (2018a) found that GREB1L is part of a larger genomic region under selection consisting of four genes on chromosome Omy28 with a major effect on maturation timing and spawning ground arrival timing. Narum et al. The place of spawning is also genetically controlled. Veale and Russello (2017a,b) found alleles associated with river-spawning and lake shore-spawning in O. nerka across their pan-Pacific distribution, involving genetic variation within, or linked to, the region surrounding the lrrc9 gene. In Lake Garda, the distinctive deep-water lake spawning behaviour of the native S. trutta results in reproductive isolation from introduced S. trutta, which are lacustrine-adfluvial spawners (Meraner & Gandolfi, 2018), emphasising that their spawning locality is genetically determined, with adaptations probably being required for spawning at such depth and pressure.

| DIRECT AND EVOLUTIONARY EFFECTS OF ENVIRONMENTAL CHANGES ON MIGRATION
All aspects associated with migration have been shown, in one or more selection may also operate where increased frequency of one life history could allow selection to favour another until a balance is achieved (Hecht et al., 2015). Thus, as the migratory fraction increases, the remaining resident fish will have reduced competition for food, which may be advantageous even where these resources are less than would have become available through migration. Similarly, the rarer male type may have a competitive advantage in spawning. These aspects are explored in more detail in Table 2.
If migration or river-residency is advantageous in particular situations, it would be expected that compensatory adaptations would occur to increase benefits relative to costs (Hendry et al., 2004). Jonsson and Jonsson (2006b) found that anadromous S. trutta body size, age at sexual maturity, relative fecundity and the ratio of fecundity to egg mass increased with distance from the sea to the spawning grounds, consistent with the hypothesis that selection favours a larger body size when migratory costs are greater. Micheletti et al. (2018b) found evidence that the environment on the migration route of migratory O. mykiss can lead to substantial divergent selection, which varied on a regional basis. Migration distance to the sea and mean annual precipitation along the route were significantly associated with adaptive genetic variation. Additional variables such as minimum water temperature during migration and mean migration elevation were significant only in long-distance migratory inland stocks. Adaptive variation associated with migratory landscape features was considerably greater than that associated with natal-site landscape features. Distance from the feeding habitat to the spawning ground is an indicator of migration costs in terms of energy expenditure and mortality in migratory fishes. The time required for migration for a given distance is also likely to be an important factor in energy expenditure and is not simply related to distance but to barriers, presence of lakes, etc.
( Table 3). However, altitude and distance of migration are negatively correlated with anadromy in S. trutta (Jonsson & Jonsson, 2006b;Ruokonen et al., 2018). The migration distance to reach the feeding destination coupled with the difficulties of the return migration (i.e., migration harshness) have been shown to have a strong effect on the bioenergetic costs involved with anadromous salmonines that migrated longer distances being more efficient in energy use than short-distance migrants (Bernatchez & Dodson, 1987 F I G U R E 4 Summary of how genetic, environmental and parental factors could interact to determine the life history of Salmo trutta and how evolutionary changes to life history could result from environmental changes that alter the relative reproductive success of migration to a particular habitat v. river residency determines migration. They found a decreased size at maturity with increasing distance from the sea; i.e., more males matured as residents. Size also covaried for the two species among 10 tributaries of a catchment covering a range of 100 km, consistent with either convergent evolution or convergent plastic responses. For natural populations above upstream impassable barriers, there is clearly strong selection against migration since migrants are lost from the population, which results in genetic differences in respect of other life-history aspects as well (Thrower et al., 2004;Thrower & Hard, 2009). As expected, in most studies of S. trutta and O. mykiss using genetic markers, there is no evidence of downstream gene flow from the above-falls populations, although in a few cases there is evidence of limited active or passive movement (Ferguson et al., 2017).
Anadromous traits may persist above barriers, despite strong natural selection against this trait because of phenotypic plasticity, or negative correlation with other traits, e.g., male maturation (Thrower et al., 2004;Phillis et al., 2016), or because some aspects of the migratory life history are selectively favoured despite the lack of access to the ocean. Once the barriers are removed or modified migration may be resumed as has been shown for several salmonid species with anadromous individuals arising from resident, fluvial-adfluvial or lacustrineadfluvial ancestors (Archer et al., in press, Quinn et al., 2017, Weigel et al., 2014; an important consideration in the restoration of extinct populations. Although the initial migrants can show poor smoltification and low marine survival (Archer et al., in press;Thrower et al., 2004), these aspects of the migratory syndrome are expected to improve over subsequent generations under the influence of natural selection.
In a common-garden experiment involving offspring of an anadromous O. mykiss population and an above falls resident population derived from it in 1910 (Scott Creek), Phillis et al. (2016) found that the frequency of age 1+ year smolts in above-barrier offspring was 54% T A B L E 2 Some non-mutually exclusive hypotheses to explain why potamodromous and anadromous migrations of Salmo trutta are facultative rather than obligate

Ecological conditions vary across time
If the relative fitness of migratory and resident individuals varies through time, temporally fluctuating selection may favour the capacity of individuals to produce either type depending on physiological condition relative to a genetic threshold or bet-hedging (where tactics develop randomly). Examples: In some years, or for some cohorts, relative growth and survival benefits downstream, in a lake or at sea may outweigh those in the natal river, but in other years, the reverse may be true. Thus, neither tactic outcompetes the other in the long-run.

Ecological conditions vary across space
If the relative fitness of migratory and resident tactics varies across habitat types within a single freely interbreeding population, this may select for individuals that are capable of producing either tactic. Examples: Fry that rear in more productive parts of the river, or that obtain better feeding territories, may be better off remaining resident and maturing early, whereas fry that rear in lower-energy environments may gain more by becoming migratory. Smaller tributaries or spawning areas with smaller gravels may select for smaller resident females, whereas larger tributaries or areas with larger gravels may favour larger migratory females. A relatively small amount of gene flow among habitats/tributaries within rivers will still be enough to prevent genetic differentiation at neutral markers and possibly also at adaptive markers underpinning migratory decisions. But even in the absence of any spatial genetic differentiation within catchments, spatial variation in ecological conditions coupled with dispersal can select for conditional strategies (Moran 1992) and thereby produce spatial variation in migratory tactics.

Frequency dependence favours a stable mix of tactics
Smaller resident males may 'sneak' more fertilisations when rare, whereas larger migratory males may obtain more fertilisations on average when small resident males are most abundant. This mechanism can act to stabilise tactic frequencies at some intermediate value or, in theory, could lead to constant cycling of tactic frequencies. Examples: Early maturing resident males have a spawning advantage relative to migratory males only when rare.

Sexually-antagonistic selection maintains genetic variation in anadromy
The evolutionary interests of males and females may be in conflict, such that genes that increase the propensity for migration are selected for in females but against in males. This then maintains genetic variation in the propensity for migration. Examples: Females carrying genes for higher condition-thresholds are more likely to be migratory, which increases their reproductive success, but their sons may then inherit these same genes and hence also become migratory, which may be less optimal for males than residency. Such 'sexual conflict' may mean that neither tactic has superior fitness overall (averaged across males and females), hence both co-exist.

Heterozygote advantage favours the maintenance of genetic variation in anadromy
For a given genetic locus affecting the propensity for migration, two or more alleles can be maintained in the population by balancing selection if heterozygotes have higher fitness than homozygotes. Example: Heterozygous parents produce a mix of migratory and resident offspring, whereas homozygous parents produce more of one type than the other. If selection on average favours some intermediate threshold for migration, heterozygotes may have a long-term fitness advantage over homozygotes. This mechanism could partially explain why genetic variation in migratory thresholds is maintained, but by itself does not explain why an intermediate degree of anadromy is favoured (although the other hypotheses might).

Optimal feeding destination may change over time
If the genes determining the migration decision and the migration destination are linked they will co-vary and not evolve independently. Example: Feeding at sea may be best at one time but in a lake at another time, thus preventing genes responsible for the decision and destination being fixed. This is a special case of Hypothesis 1 above, with the added twist of genetic trade-offs among traits (migration decision versus destination) due to pleiotropy (where the same genes affect multiple traits). Theoretical considerations, however, suggest that antagonistic pleiotropy may maintain genetic polymorphism only under a rather restrictive range of conditions (Hedrick, 1999), so this mechanism may be less important relative to the others.
compared with 73% in below-barrier offspring and significantly more below-barrier smolts were detected moving downstream compared with above-barrier smolts. Seawater trials showed a 37% relative reduction in salinity tolerance in the above-barrier offspring. Mature males accounted for 27.8% of all above-barrier males but only 5.4% of belowbarrier males. These changes are consistent with natural selection for river residency in the c. 25 generations since establishment.
Sahashi and Morita (2018) examined how the migratory threshold size changed in response to opposing effects of natural and artificial selection in facultatively migratory male O. masou. In fish from above an impassable waterfall it was found that, in this high-cost migration situation, the size threshold for migration had changed in the direction that promoted residency, relative to that in the below-falls population.
By contrast in the obligatorily resident S. malma, the size threshold did T A B L E 3 Anthropogenic factors potentially resulting in fitness changes and thus alterations to the cost-benefits of migration v. residency or migration destination in Salmo trutta and other salmonines

Factor Impact on migrants References
Partial barriers to downstream and upstream migration resulting from water offtake, hydroelectric generation, etc.
Increased energy expenditure. Increased risk of predation. Migration speed of smolts significantly slower. High downstream passage mortality of S. trutta kelts at hydropower stations. Upstream may be size selective and thus change size-age at maturity. Multiple partial barriers have an effect equivalent to an impassable barrier. Partial barriers resulting in a reduction of MAR alleles in Oncorhynchus mykiss. Removal of six partially impassable weirs in a Danish river resulted in nine-fold increase in spawning S. trutta over 12 year period. Apgar et al., 2017;Buddendorf et al., 2019;Haugen, 2008;Huusko et al., 2017;Jepsen et al., 1998;Ostergren & Rivinoja, 2008;Van Puijenbroek et al., 2018; Complete barrier to upstream migration resulting from; e.g., construction of water storage reservoirs and hydropower stations without fish passes.
Anadromous populations extinct. Most of 72 anadromous S. trutta populations in Finland now lost. Change in destination; e.g., anadromous become lacustrine-adfluvial migrants.
Un-naturally high and low flows resulting in decrease in or elimination of migrants. Delays and increased energy expenditure. Changes in speed of migration. Fluvialadfluvial became river-resident due to reduced habitat quality. Increased predation by piscivorous birds and mammals in downstream sections of rivers, in lakes, and at sea.
Reduced survival. Increased energy expenditure in predator avoidance. Greater increase in predation at sea tips balance in favour of potamodromy. Predation through lakes and on sea entry main factor determining number of returning anadromous S. trutta in Denmark. High predation by great cormorants Phalacrocorax carbo key mortality factor in some rivers & lakes. Heavy pike Esox lucius predation at river-to-lake confluences. Berejikian et al., 2016;Healy et al., 2017;Jepsen et al., 2018Jepsen et al., , 2019;Kennedy et al., 2018;Schwinn et al., 2018 Increased exploitation. Differential life history, size, and sex exploitation.
Reduced marine survival due to exploitation either directly or as a by-product. Greater exploitation of (larger) migrants than (smaller) river-residents resulting in selection for latter. Selection for earlier age of maturity, run timing and time of spawning. Czorlich et al., 2018;Hollins et al., 2018;Kallio-Nyberg et al., 2018;Koeck et al., 2018;Syrjanen et al., 2018;Thériault et al., 2008;Tillotson & Quinn, 2018 Climate change. Changes in river flows and water temperature influencing feeding, migration timing, spawning and juvenile survival. Increased metabolic cost of upstream migration. Decreased marine productivity and increased freshwater productivity and growth rates tipping balance in favour of potamodromy-river-residency. Possibly direct effect of temperature on life history. Interbreeding with stocked fertile hatchery reared / farm S. trutta.
Decreased genetic tendency for migration. Reduced marine cf. freshwater survival. Ferguson, 2007;Ferguson et al., 2017;Thrower & Hard, 2009; MAR: migration-associated region. not differ in above and below-waterfall populations indicating that environmental differences did not affect it. In a hatchery strain of O. masou that was subject to artificial selection for migration, the threshold altered in a way that favoured migration.
The numbers of returning anadromous S. trutta has declined over recent decades in many parts of north-western Europe Rasmussen et al., 2019). Similarly, lacustrine-adfluvial S. trutta have declined in some countries including Finland (Syrjanen et al., 2018) and Switzerland (Gafner & Meyer, 2018). The sustained nature of these declines means that genetic changes have probably occurred in response to changes in fitness and thus natural selection.

for anadromous
O. mykiss suggest that if sea-survival rates are reduced by some 50%, anadromy no longer occurs, although the extent of the reduction in survival required was population specific .
The decrease in anadromous S. trutta numbers has involved multiple anthropocentric factors during both out and return migrations, as well as at the feeding destinations, which have resulted in both a reduction in survival of migrants or an increase in the costs of migration (Table 3). Thus, although, for example, many barriers do not stop downstream and upstream migration occurring there is cumulative increase in the likelihood of predation and a cumulative energy cost due to delays. The benefits of migration are reduced and the balance is potentially tipped in favour of river-residency. This can also result in a change to the migration destination, for example, anadromous S. trutta become lacustrine-adfluvial migrants.

| FURTHER GENETIC STUDIES
Studies on the genetic determinants of migration in S. trutta lag substantially behind those on other salmonines especially O. mykiss, although even for the latter species most studies have, until recently (Arostegui et al., 2019), only involved anadromy (Kendall et al., 2015).
With current rapidly changing environmental conditions and diminishing numbers of individuals undertaking migratory life histories, studies of such determinants in S. trutta are urgently required.
The development of genomic techniques now makes this feasible.
Given the considerable similarities of facultative migration in the two species, O. mykiss genomic studies could act as a springboard enabling rapid progress to be made in respect of S. trutta.
Much more attention is required to be given to potamodromous migrations. Just as cross-taxa comparisons can be informative (Dingle, 2006;Sahashi & Morita, 2013), so comparative studies of different migration patterns within S. trutta may be more informative than focusing on a single life history such as anadromy. It is emphasised that studies need to target early developmental stages, as this is when the migration-residency decision occurs and not focus exclusively on later stages such as smolts. This will require making use of experimental lines derived from river-resident populations and populations with a high incidence of migration, together with innovative approaches, e.g., using associated genetic markers to predict the future life histories of individual fish at an early developmental stage. Studies need to be undertaken on all aspects of migration through to spawning of returning migrants. Sex needs to be taken into account in all studies, using genetic sex determination methods where necessary for juveniles.
Ecological, behavioural and physiological studies of the propensity for migration have often being carried out against variable genetic backgrounds where both intra-population and inter-population genetic variability was present, often with conflicting results. It is generally well recognised that examining genetic differences between populations or other groups requires studies to be carried out in communal environmental conditions (common-garden experiments), with reciprocal hybrids to control for parental effects. However, it seems less widely appreciated that investigating the influence of varying