Project Topics
Get Complete Project Material File(s) Now! »
How partial migration is maintained over evolutionary times
Migration depends on habitat suitability and evolution (Ronce et al., 2000). The fate of species out of their native range is either adaptation or disappearance. If introduced in an unsuitable environ-ment, individuals might need to disperse, which implies migration and reproduction, to optimize their tness. Thus, in newly colonized habitat, individual dispersal propensity is expected to evolve. Both proximate and ultimate processes are involved in the migratory choice. Proximate mechanisms act on short timescales by direct changes in the environment, they are therefore involved in plasticity, while ultimate mechanisms act over longer periods by genetic changes. Although plasticity can also evolve, and hence be impacted over longer timeframes by ultimate processes. The decision to migrate is dually controlled by an environmental/genetic threshold that determines whether an individual may migrate or not (Chapman et al., 2011b). However, disentangling the e ects of proximate and ultimate mechanisms underlying partial migration is complex. A better understanding of the mechanisms underlying migra-tion will serve to improve our knowledge on its ecological and evolutionary consequences.
Environmental instability may a ect the ability of organisms to reach their optimum in terms of reproduction and survival (Kuno, 1981). One way to face it is to disperse. As the rate of disturbances increases, the dispersal ability increases. Dispersers may reach climate refuges (Morelli et al., 2016) or new habitats that become favourable arising from climate change. Those habitats give new colonization opportunities for dispersing species (Lawler, 2009). The Arctic and Antarctic regions will soon turn into suitable ecological niches for new species. In a context of global change, the instability may arise from a change in meteorological conditions (e.g. temperatures, precipitation, river ow rates) or from a change in the community structure (e.g. population density, trophic chains) (Brodersen et al., 2011). Temperature is known as an environmental factor governing numerous developmental and life stages (e.g. plant development (Bollero et al., 1996), migration strategy (Alonso et al., 2009; Nilsson et al., 2006; Skov et al., 2010 ), and seasonal timing of life cycle events (Cagnacci et al., 2011; Gienapp et al., 2014; Reed et al., 2010)). Diverse other environmental conditions may a ect the individual’s behaviour such as rainfall (Boyle, 2011), res and vegetation biomass (Naidoo et al., 2012). When population size increases, density is expected to become the main driving force of dispersal (Grayson and Wilbur, 2009; Kaitala et al., 1993), acting on the competition for habitats, resources or reproduction (Loe et al., 2009; Nathan et al., 2008). On the contrary, when the population size is small, kin competition is the selective force driving the evolution of dispersal (Ronce et al., 2000). Under the hypothesis of non-saturated habitat under competition, a decrease in the dispersal ability is observable when the survival of migrants is lower than residents’ survival. If migrants have a lower fecundity than residents (number of o spring produced in a site), the dominance of the resident strategy is expected in such case. At the individual level, dispersal may be heritable. The heritability may be direct such as the herit-ability of behaviour (Paez et al., 2011). Phenotypically dimorphic traits have been demonstrated to be partly heritable (Ro , 1998), such as the age at migration in some shes or the seasonal migration in birds (Pulido and Widmer, 2005). The heritability may also be indirect depending on the energetic status of breeders or during the development of juveniles (e.g. variation in movement capacity or competitive ability linked to some traits such as body size or weight). For example, the food restriction at a certain developmental stage has a signi cant e ect on the migratory tactics, also correlated to sexes (Archer, Hutton, Harman, O’Grady et al., 2019). In term of behaviour, dispersal may happen before the rst breeding attempt (natal dispersal) or between two consecutive breeding attempts (breeding dispersal) (Harts et al., 2016). In the case of natal dispersal, stable and fragmented habitats may create dispersers as a bet-hedging strategy (Clobert et al., 2012). For example, the seeds dimorphism in achenes leads to di erent strategies of dispersal through space and time. The peripheral seeds have higher dormancy and lower dispersal abilities than the seeds from the centre of the capitulum. This bet-hedging strategy reduces the impact of spatio-temporal changes on the reproductive ability of the species (Imbert, 1999). In term of development, body size is often a good proxy of the decision to disperse (Alerstam et al., 2003; Gyllenberg et al., 2008; Jahn et al., 2010). Divergence in body size between individuals may arise either from phenotypic plasticity and/or genetic variation (Rogell et al., 2013).
Di erent hypothesized factors act as drivers of dispersal. First, the competitive release hypothesis, where larger individuals outcompete the smallest, and hence the smaller subordinates disperse to achieve their needs (Yeaton and Cody, 1974). When the density increase, there should be an increase in the number of migrants. This can be true in populations where predation risks is low, on the contrary, when it is high, individuals may bene t more from living together and protecting the population as a group of residents (Chapman et al., 2011b). Second, the arrival time hypothesis, simply summarized by rst arrived rst served. Individuals with high phenotypic quality often arrive rst on sites, this is the prior residency e ect (Kokko et al., 2006). Furthermore, the dynamic of one tactic seems to depend upon the overall population demography (Griswold et al., 2011). Kokko (2011) has even shown that rules of territ-ory acquisition matters, with either residents or migrants becoming predominant in population just under the arrival time hypothesis. Thus, partial migration could be also explained by a demographic balan-cing at the population-level (Hebblewhite and Merrill, 2011). Under this hypothesis, smaller individuals are expected to be less competitive in territory acquisition. This theory is prone to be sex-dependent, where males often larger than females are prone to residency and territory defence. Thirdly, the fasting endurance hypothesis (Millar and Hickling, 1990), is based upon the individual physiological di erences in survival. Now it is more frequent to talk about the ‘limited foraging opportunity’ hypothesis, where individuals who do not tolerate the change in food intake may migrate (Boyle, 2011). Of course, spatial and temporal distributions of food may be unequal and can play a role in partial migration. The ‘in-creasing food availability hypothesis’ is also suggested by Gross (1987). Migration thus occurs when the food balance is di erent between habitat of departure and arrival (Ayer et al., 2018).
Fish caught during the long-term monitoring in the Kerguelen Islands
Long-term monitoring in the Kerguelen Islands began in 1955 with rst introductions of eigth sal-monids species (Coho salmon (Oncorhynchus kisutch), Rainbow trout (O. mykiss), Chinook salmon (O. tshawytscha), Arctic char (Salvelinus alpinus), Brook trout (S. fontinalis), Lake trout (S. namaycush), Atlantic salmon (Salmo salar) and Brown trout (S. trutta)). After 22 introduction attempts and 23 rivers seeded (Labonne et al., 2013; Lecomte et al., 2013), natural colonization was achieved by anadromous individuals. Among the eight species introduced, three failed to establish local population (rainbow trout, chinook salmon and lake trout). Since 1962, 81184 shing data were collected as recorded in Table 3.1. The database used in this work is still being updated. Therefore, not all sh caught are represented in the database we have been working on. The last update considered in this work dates from the end of 2019. When referring to » sh caught », « sampling » or « available data », we mean implemented data in this database.
The complete history of the salmonids introduction in the Kerguelen Islands is recorded in Lecomte et al., 2013 (Fig. 3.1). The major introductions of brown trout (mainly originated from Bidarray in the French Pyrenees), brook trout and rainbow trout occurred from 1955 to 1962. From 1975 to 1981, Atlantic salmon, brown trout (originated from the Baltic Sea, River Slupia in Poland) and rainbow trout were released under control. Meanwhile a transfer experiment was launched to study the colonization process along the coast (reproduction, colonization, and speed of the process). The ten next following years were devoted to a sea ranching experiment of chinook salmon, brown trout, brook trout and arctic char (originated from Lake Leman, France) in Armor. In 1997, 33 rivers were recorded as colonized. In 2019, natural colonization was proven only for brown trout, brook trout, Arctic char and Coho salmon. The brown trout represents a major part of the database with 74%, followed by brook trout and Atlantic salmon. This number of data does not re ect the real number of available sh. The explanation stands in the capture-mark-recapture campaign (CMR) that was lead during the 70’s. The idea was to mark sh with tags to follow their migrating movement. The sh belonging to that program were caught regularly. Their size and phenotype-at-capture (resident/migrant) were noted, and scales were removed for ageing and growth analysis. According to their movement, some sh were caught several times, so that one line in the database is representative to one capture event of sh and not representative of one sh. The available number of sh is approaching 70000. It is important to note that the database used in the present work is the one updated in 2019. Because of an important work of implementation of the database made by F. Gueraud, the status of the database changes rapidly. This work is in progress. For information, more than 151000 shing data are available at present.
All the introduced species in the Kerguelen Islands belongs to the Salmoninae subfamily of the Salmonidae. They all originate from the Northern hemisphere and reproduce in cold fresh waters either one (semelparity, most Oncorhyncus) or several time (iteroparity, Salmo and Salvelinus). After eggs deposition under the appropriate substrate, the fry emerges and grow in fresh water at juvenile stage.
The brown trout in the database
More than 74% of the sh in the database are brown trout (60471 individuals) (Table 3.3). They represent on average 70% of the collected samples, scales, or otoliths sampling. Less than 3% of the sh in the database were sacri ced for otoliths studies. In the case of the brown trout, the sampling was done in 62 di erent locations in the Kerguelen Islands. The last shing campaign was led from Decem-ber 2018 until February 2019. For further information on the last sampling in the eld, see Appendix A.2.
To characterize the life-history traits of sh and colonization dynamic of salmonids, sh sampling was mainly performed to collect calci ed structure of sh (scales and otoliths). Scales were sampled in the optimal zone, located near the dorsal n above the lateral line (Elliott and Chambers, 1996). Method of shing, localization (rivers, coordinates) and date are recorded. Body-size (fork length in millimetre (mm)), weight in gram (g), and species are noted. The phenotype at capture is also de ned when possible. As partially migratory species, resident and migrant brown trout di er by their phenotype. They are di erentiated by morphological traits (i.e. length/weight relationship (Jonsson, 1985), silver-grey colouration (Quigley et al., 2006)). Additional information was taken if available, such as sex, or health state. Depending on the study, other samples are taken, some non-lethal such as n sampling, some lethal such as otoliths or stomach content. Investigations were conducted according to the guiding principles for the use and care of laboratory animals and in compliance with French and European regulations on animal welfare (Decret 2001-464, 29 May 2001 and Directive 2010/63/EU, respectively).
How to select the sh in the database: description of the decision rule
To study the spatio-temporal variation of life-history traits taking data dispersal into account, we need to select sh in a spatio-temporally comparable framework. Along a temporal continuum ranging from 1962 until now, sh must be chosen in the same river (spatial e ect) and the same cohort (temporal e ect, where cohort is de ned as groups of individuals born the same year). By doing so, we will have sh experiencing comparable environment of growth, as we expect density of population and temperature to change through time. To select sh within cohorts, we need to infer their probable age knowing their size. To do so, scales are often used to access the life-history traits, by counting the annuli deposited yearly in the structure (Beall and Davaine, 1988). Nevertheless, using scalimetry to empirically determ-ine the individual age of more than 60 000 samples can be time-consuming. As already mentioned, the long-term monitoring in the Kerguelen Islands gives access to a huge dataset and the allocation of time to the scalimetry procedure does not allow studying all of them.
Another solution to infer the age of sh is to establish a relationship between the age and body size at capture based on aged sh. On the 28968 Brown trout in our dataset (measurement of fork length at capture and collection of scales) (see Table 3.3), 6775 sh have already been aged through other studies (23%) lead in the Kerguelen Islands. The already aged sh in the database belongs to 7 rivers, Rivers Ac na, Albatros, Ch^ateau, Nord, Norvegienne, Studer, and Val-Travers, some common with our study and others. The relationship between age and body size at capture can be estimated from those samples. 22193 sh remain to be aged. According to the phenotype at capture, migrant are bigger than resident, and numerous sh were not attributed a phenotype when captured (Figure 3.8). The majority of the sh caught belongs to a size category inferior to 400 mm (Fig. 3.8a) but all the size classes are represented among the aged sh (Fig. 3.8b).
Nevertheless, the relationship between the age and body size is also thought to depend upon the phenotype at capture. The expected increase in body size depends upon conditions and availability of resources in the environment in which the sh is growing (Davaine and Beall, 1992). As the brown trout is a facultative anadromous species, we can expect resident to grow slower than migrant at least out of the reproductive period (when migrants are at sea). Even if pristine fresh water habitats in the Kerguelen Islands may allow good access to the resources, sh growth at sea is expected to be better (Beall, 1979; Jarry et al., 2018. The size classes distribution of phenotypes (Figure 3.8) is congruent with this hypothesis. The non-identi ed phenotype (NA) is mainly represented in the lowest size classes, nevertheless a non-negligible part of the samples belong to the highest size category (superior to 400 mm). In the database, the phenotype is described by 115 levels. The rst selection rule was to homogenize the phenotype in 4 categories: resident trout (RT), migrant trout (MT), lake trout and non-identi ed (NA). As lake trout are not consistent enough in the sampling, we considered them as resident trout to reassign an age to the non-aged sh. The smolting individuals had size ranging from 94 to 667 mm. We decided to reassign smolting individuals to the category of migrant trout when their size was superior to 200 mm. If not, as for the remaining sh, the phenotype was non-identi ed. The number of sh (n) in each category is summarized in Table 3.5.
The objective of this part is to de ne a methodology aiming at redistributing an age to the sh that have not yet been aged through scalimetry. We alternatively chose a method called out-of-bag error, also called out-of-estimate. This method is used to measure the prediction error (PE), where PE (or also accuracy) is a measure of how close model predictions are to their true value on average (De’ath, 2007). Di erent techniques are used to evaluate models’ performance, such as random forest, averaging, shrinkage, bagging or boosting trees. The principle of random forest is to bootstrap sub-sample of the data set to create training set to t the model (multiple trees) with a random subset of predictors (Prasad et al., 2006). This step is repeated a high number of times. For our analyses, we used 20000 bootstraps. The training sample was used to reassign an age to the bootstrapped samples according to the decision rule. We tested for two decision rules to calculate a probability to belong to a discrete age knowing the size at capture.
From the back-calculation of size at age to the estimation of individual growth
Based on the variance decomposition work and because scales are estimators of life-history traits, in this work we used four scales to determine the age of sh (age at migration and total age) and considered a single scale for associated growth measures. The scales were prepared and read according to the same methodology as previously described, except for double reading by di erent readers. All measures of growth (position of annulus) and the same variables (total age TA, freshwater age FA, total radius TR, freshwater radius FR) were noted for all the scales making up our data set.
In many species, the growth of calci ed structures is correlated with somatic growth (Casselman, 1990; Gunther et al., 2012; Ottaway, 1978). When the size of the sh at each age cannot be obtained from CMR-type monitoring, the growth recorded on the scales, whether daily or annual, makes it pos-sible to rebuild the probable size of the sh at a given age. The use of calci ed structures to infer the size of sh at a previous age is called back-calculation (Casselman, 1990; Iba~nez et al., 2008; Kipling, 1962). A back-calculation model is used to estimate the length of the sh at age i : Li, from known values of length at capture (total length, LT ), and measurements made on scales : total radius (TR or RT ) and distance between the core and the radius at age i : Ri. Numerous back-calculation models exist (Ogle, 2013). Models are described hereafter, and the main mathematical equations are recorded in Table 4.7. Widely used back-calculations methods are the one of Dahl and Lea (1910), Fraser-Lee (1916-1920), the body proportional hypothesis (BPH, also called Whitney and Carlander) and the scale proportional hypothesis (SPH, also called Hile). Di erent assumptions are underlying all these models. Dahl and Lea suggest that scales growth is an exact proportion of sh growth (1). Fraser-Lee implemented the previous model by allowing sh to grow before the formation of the rst scales (2). The SPH assumes that if a scale is bigger than the average scale for that size of sh (catchment size), then this di erence would be conserved throughout the lifespan (3). The BPH is the exact same thought but turned the other way round. Meaning that if a sh is x percent smaller or bigger at capture than the average sh for that size of scale, then this di erence would be conserved throughout the lifespan (4). Other numerous models exist such as nonlinear SPH, nonlinear BPH but the preferred back-calculations model is often the Fraser Lee model (Duncan, 1980; Kouhestan Eskandari et al., 2018). However, it is interesting to note that studies are carried out to choose the best back-calculation model in accordance with the study species or the calci ed part chosen (Horppila and Nyberg, 1999; Iba~nez et al., 2008).
Table of contents :
I General introduction
1 Conceptual framework
1.1 Ecology and evolution of life-history traits
1.2 Migration one key to life history and dispersal evolution
1.3 How partial migration is maintained over evolutionary times
1.4 What are the consequences of partial migration in the evolutionary process?
2 Study framework
2.1 The brown trout (Salmo trutta, L. 1758)
2.2 Objectives of the dissertation
II From sh and scales sampling to the implementation of an analytical methodology
Objectives and organization of the part
3 Individual selection in a spatio-temporal framework
3.1 Fish caught during the long-term monitoring in the Kerguelen Islands
3.2 The brown trout in the database
3.3 How to select the sh in the database: description of the decision rule
3.4 Conclusion .
4 The use of scales to reconstruct life histories
4.1 The general use of calcied structures
4.2 Hierarchical variance decomposition of sh scale growth and age to investigate the relative contributions of readers and scales
4.3 From the back-calculation of size at age to the estimation of individual growth
4.4 Conclusion .
Discussion on the sampling design and the chosen methodology
Bibliography
III … To study the evolution of life-history traits
5 The evolution of growth and body size at age
5.1 Growth rate or body size? How to model the variable driving migration?
5.2 On the evolution of freshwater growth prior to migration
5.3 The evolution of body size of partial migratory population on shifting expanding range
5.4 Conclusion .
6 Towards an evolution of the migratory pattern?
6.1 The link between growth and migration, the reaction norm principle
6.2 Towards the evolution of the threshold size at migration
IV Conclusive part
7 General discussion
7.1 Quantifying the variation in life-history traits
7.2 Mechanisms promoting dispersal
7.3 The context of expanding populations
8 Conclusion and perspectives
V APPENDICES I
A Supportive information to Chapter 3. Individual selection in a spatio-temporal framework III
A.1 Status of knowledge in 2019 on presence of salmonids in streams and waterbodies of the Kerguelen Islands .
A.2 Campagne de terrain 2018-2019 dans les Terres Australes et Antarctiques Francaises
A.3 Microchemical analysis of Manchots lagoon
A.4 Number of samples available in each phenotype category
B Supportive information to Chapter 4. The use of scales to rebuilt life histories
B.1 Hierarchical variance decomposition of sh scale growth and age to investigate the relative
contributions of readers and scales
C Supportive information to Chapter 5. The evolution of growth and body size at age
C.1 Modelisation de la croissance. Resultats preliminaires
C.2 Convergence des cha^nes du modele de Von Bertalany
C.3 Estimates of body size at age by rivers
C.4 Selection of age-1 sh with Neural Network Learning
D Activites scientiques
