Evaluation of methods
The use of stable caesium (133Cs) as a tracer for estimating brown trout food consumption appeared to be very successful. With the exception of age-group 0+, the estimated daily food rations were similar to predictions from a laboratory-based model for maximum consumption (Elliott 1975a,b). For the smallest fish (0+), the food rations were much higher than the maximum rations. This may be expected as Elliott (1975a,b) never used fish this small and extrapolations from larger fish may be invalid. The estimated daily rations matched the general expectation for the effect of temperature, body size and season. Daily rations were higher during the first sampling period (July–August) than during the second period (August–September), when temperature was 2·7 °C lower and food abundance was probably lower as, e.g. large insect larvae have hatched. In accordance with expectations, daily mass specific rations declined by age and body size (e.g. Wootton 1990).
To our knowledge, stable caesium as a tracer element has been used to estimate food consumption in fish on one occasion only (Hakonson, Gallegos & Whicker 1975). Forseth et al. (1992) compared estimates for brown trout food consumption based on the turnover of radioactive caesium (137Cs) from the Chernobyl fallout, with estimates from a well established method based on the amount of food in the stomach and the rate of gastric evacuation (Elliott 1972; Eggers 1977). The estimates were very similar, and Forseth et al. (1992) concluded that the use of caesium as a tracer is a reliable method for estimating food consumption. In principle, there is no difference between using stable and radioactive caesium as a trace element.
The high food consumption among 2+ migrants and the large difference in consumption rate between migrants and the stream dwellers is essential to our conclusions. As we were unable to differentiate between migrating brown trout and those that remained in the stream before the fish actually started to move downstream, we had to estimate the initial caesium concentration (on 12 July) of the migratory individuals. However, the estimated daily ration was essentially insensitive to variation in initial caesium concentration. A 100% increase in concentration 1 month before the downstream migration started caused only a 4% reduction in the estimated food ration. The main reason for the much higher food consumption among migrants than those that remained in the stream is that migrants had 1·5 times higher caesium concentration when they were caught in the trap. The estimated food consumption for migrants was nearly four times higher than maximum consumption predicted from laboratory studies (Elliott 1975a,b). This is very high and indicates that fish leaving the river at an age of 2 years are feeding at an exceptionally high rate, making them energy-wise the best performing fish in the population.
Evaluation of hypothesis
In accordance with our hypothesis, fast-growing brown trout migrated earlier and at smaller body size than slower-growing individuals. 2+ migrants were significantly larger than those that remained in the stream, and 3+ migrants were significantly larger than 2+ migrants. Comparisons across cohort (i.e. 2 and 3+ migrants) may be questionable because cohorts may experience different growth rates (for example due to variable year-class strength and density-dependent growth), but judging from the magnitude of the size difference (3+ being nearly twice as heavy as 2+), it is obvious that the young migrants were smaller than those that migrated 1 year older. This result appears general among migratory salmonids as it accords with findings from studies on brown trout (Bohlin, Dellefors & Faremo 1993; Jonsson & Gravem 1985; Jonsson 1985; Bohlin et al. 1996), Atlantic salmon (Jonsson et al. 1990; Økland et al. 1993), sockeye salmon (Oncorhynchus nerka) (Burgner 1991) and Arctic charr (Salvelinus alpinus) (Forseth et al. 1994).
The food consumption and energy budgets were much higher for migratory than stream resident trout. The absolute daily ration for 2+ migrants was more than four times higher and the energy budget (i.e. the energy of consumed food) 4·5 times higher than for resident 2+ fish. Despite this large difference in food consumption, the specific growth rate did not differ significantly between resident and migratory individuals. However, the total energy allocated to growth, and thus their mass increase, was higher among migrants. Moreover, the total metabolic costs were five times higher among migrants than among resident fish. In the present study, it was impossible to differentiate between the different components of metabolic costs. A large proportion of the estimated difference can probably be explained by the higher costs of digesting and assimilating (specific dynamic action, Rf in eqn 3) a much larger amount of food for migratory than resident fish. However, as the differences in metabolic costs between the two groups were larger than the differences in energy accumulated through food, the hypothesis of higher standard metabolic rates among early migrants of Atlantic salmon (Økland et al. 1993; Metcalfe et al. 1995) is given some support. An alternative explanation is that migrants have higher activity costs than those that remained in the stream. The distribution of metabolic cost among the components is, however, not essential to the conclusions in the present study.
Although the total energy allocated to growth and the mass increase was higher among migrants, their proportional allocation to growth was much lower than that of resident fish (about half). All 2+ brown trout had lower proportional allocation to growth than 1+ fish, but the reduction was larger among migrants (88%) than among resident fish (68%). A reduction in the proportional energy available for growth is a likely explanation for why migration is initiated at age 2. Moreover, it may explain why some individuals, those with higher metabolic rates, migrate earlier than others. They experience a larger drop in their proportional energy available for growth and seek alternative actions, such as migrating to an alternative feeding niche, to maintain their status as fast growers. Slower-growing individuals experience a smaller drop in proportional energy available for growth and remain in the stream for one more year. Most 3+ individuals migrate to the lake, but some of the largest males mature sexually and remain in the stream.
Migratory costs for juveniles in Litjåa can probably be neglected as the migrations are short and no change in salinity occurs. Thus, the optimal time for migration is the one that maximizes the ratio between the growth benefit of changing habitat and the costs of increased mortality after migration. Post-migration mortality of salmonids is often assumed to be negatively size-dependent because predation is size-dependent; small migrants are susceptible to a higher number of predatory species and a wider size range of predators than larger ones (Bohlin et al. 1993, 1996). As young migrants are smaller than older migrants, early migration is more risky than late migration.
Brown trout cannot, prior to migration, measure the growth they will attain in a new habitat. To optimize the time of migration, individuals thus have to respond to some change in conditions in their present habitat. The underlying mechanism which supports the ‘decision’ to migrate is assumed to be related to growth rate, or a physiological process like metabolic rate which is correlated with growth rate (Jonsson & Jonsson 1993). Brown trout from Litjåa experienced a relatively large drop in growth rate from age 1+ to 2+ , but migrants maintained as high growth rates in the summer of migration as resident fish. Thus, the growth rate per se cannot explain why some individuals migrate earlier than others. However, their relative allocation to growth declined significantly (from age 1+ to 2+), and young migrants experienced a larger drop than older ones. Thus, it appears that the fish are able to measure, by some physiological mechanism, changes in their amount of surplus energy available for growth, as postulated by Thorpe (1986).
Juvenile brown trout thus appear to migrate from one habitat to another as a phenotypically plastic response to declining growth performance as they reach an environmental threshold in their present habitat. This accords with the general assumption that migration is a biological response to adversity (Taylor & Taylor 1977). Individuals may reach this threshold at different ages and sizes depending on their metabolic status. Fast-growing individuals migrate earlier and at a smaller body size than slower-growing individuals, because their metabolic rates are higher, and consequently experience a larger drop in their allocation of energy to growth. By migrating, the fish are probably able to retain a higher growth rate than possible under the feeding opportunities in the original habitat.
For fast-growing individuals, an alternative to migration is to mature sexually in the stream. The size advantage attained in the stream, relative to slower-growing individuals, may then be converted into a fitness advantage by earlier reproduction and the possibly of participating in more spawning events during life. Among brown trout in Litjåa, this tactic was followed by a small proportion of the males only. These males were among the largest within their cohorts. Among females, the fitness gained by migrating to the lake and returning as large spawners with high fecundity appears to be more than balanced by the higher risk of mortality by postponing maturation. Among males, alternative mating strategies such as sneaking, may promote early maturation among fast-growing individuals, as fitness may be high for both small and large individuals (Jonsson 1985). For fast-growing males it thus appears to be alternative strategies of migration or early maturation. Most follow the first route, but some use the alternative.
All in all, the present study supports the hypothesis that fast-growing individuals shift niche earlier and at a smaller body size than slower-growing individuals, because they maintain higher metabolic rates and are energetically constrained younger than slow growers by limited food resources (Jonsson & Jonsson 1993). The sources of the variation in metabolic rates among individuals are unknown, but maternal and developmental effects and genetic diversity may all cause such variation in metabolic rates. Egg size, time of hatching and emergence from the gravel (Metcalfe & Thorpe 1992), early developmental effects and even random effects (e.g. spatial and temporal variation in the quality and availability of food items at first exogenous feeding) giving some individuals a head-start in life, may cause differentiated metabolic rates within one cohort. Genotypic differences in metabolic rates may also be maintained within a population due to variable selection pressures. A high metabolic rate is advantageous only if an additional energy intake can be attained through food consumption (Forseth et al. 1994). An individual's possibility to attain such an additional intake may depend upon cohort or population size (density dependency) and environmental factors influencing prey availability. It is important to understand the sources of variation in metabolic rates, because the consequences may be many. Recent studies indicate that in salmonids early metabolic rates represent an important premise for several of the life-history decisions the fish has to make later in life (Metcalfe et al. 1989; Metcalfe 1991; Titus & Mosegaard 1991; Metcalfe et al. 1992; Forseth et al. 1994; Metcalfe et al. 1995), and the present study shows that the metabolic status is also important for the timing of juvenile migrations in salmonids.