1 Experimental data for maximum growth and food consumption of Atlantic Salmon (Salmo salar L.) parr from five Norwegian rivers situated between 59 and 70°N were analysed and modelled. The growth and feeding models were also applied to groups of Atlantic Salmon growing and feeding at rates below the maximum. The data were fitted to the Ratkowsky model, originally developed for bacterial growth.
2 The rates of growth and food consumption varied significantly among populations but the variation appeared unrelated to thermal conditions in the river of population origins. No correlation was found between the thermal conditions and limits for growth, thermal growth optima or maximum growth, and hypotheses of population-specific thermal adaptation were not supported. Estimated optimum temperatures for growth were between 16 and 20 °C.
3 Model parameter estimates differed among growth-groups in that maximum growth and the performance breadth decreased from fast to slow growing individuals. The optimum temperature for growth did not change with growth rate.
4 The model for food consumption (expressed in energy terms) peaked at 19–21 °C, which is only slightly higher than the optimal temperature for growth. Growth appeared directly related to food consumption. Consumption was initiated ≈2 °C below the lower temperature for growth and terminated ≈1·5 °C above the upper critical temperature for growth. Model parameter estimates for consumption differed among growth-groups in a manner similar to the growth models.
5 By combining the growth and consumption models, growth efficiencies were estimated. The maximum efficiencies were high, 42–58%, and higher in rivers offering hostile than benign feeding and growth opportunities.
Because fish are obligate ectotherms (with a few exceptions), ambient temperature has a pervasive controlling effect on their rate of growth and food consumption (Wootton 1998). It is therefore important to establish relationships between water temperature, food consumption and growth. By modelling the performance under optimal experimental conditions, one can explore the reasons for deviations from this baseline in natural populations (Elliott, Hurley & Fryer 1995; Forseth et al. 2001). During the past two decades, numerous studies have applied growth models for studies of animal performance (review by Elliott 1994). One important assumption for such studies usually has been that there is no intraspecific variation for the baseline trait (e.g. Ugedal, Forseth & Jonsson 1997). If such variation occurs, the applicability of such models will change.
Recently, the old models (Elliott 1975a,b) used for salmonid growth and feeding performance were revised (Elliott et al. 1995; Elliott & Hurley 1998) and applied to Atlantic Salmon (Salmo salar L.) by Elliott & Hurley (1997) and Forseth et al. (2001) and to Arctic Charr (Salvelinus alpinus L.) by Larsson & Berglund (1998). The new growth model has a triangular shape with a sharp peak at the optimal temperature for growth. Recently, however, Forseth et al. (2001) used an alternative model, termed the Ratkowsky model (Ratkowsky et al. 1983), which gave a better fit when modelling growth and food consumption of a Norwegian population of Atlantic Salmon. There was no peaked maximum, rather the curve was quite flat around the maximum. Because the two models share a common set of parameters with biological meaning, they can both be used in inter- and intraspecific comparisons of performance in growth and food consumption.
For salmonids, there is an open question whether thermal performance varies among conspecific populations or not. One possibility is that populations are locally adapted so that natural selection shifts the temperature of optimal performance to match the prevailing temperatures of the new thermal niche (Allen 1985). There is some support for this hypothesis from invertebrates (Lonsdale & Levinton 1985; Bennett & Lenski 1993), but not from vertebrates. An alternative counter gradient variation hypothesis (Levins 1969; reviewed by Conover & Schultz 1995) suggests that populations in apparently hostile environments (low temperature, short season for growth, strong competition for food) perform better at all temperatures than conspecifics from benign environments. This hypothesis has been supported by some studies on invertebrates, but also studies on vertebrates such as green frogs (Berven, Gill & Smith-Gill 1979), Atlantic Silverside (Conover & Present 1990) and Atlantic Salmon (Nicieza, Reyes-Gavilán & Braña 1994). For salmonids, however, Elliott (1994) questioned the existence of such local adaptation in growth performance. From this group, there is no conclusive evidence that population-specific variation in thermal performance has a genetic basis. In a study of 42 European (from 44 to 70°N) populations of Brown Trout (Salmo trutta L.), Jensen, Forseth & Johnsen (2000) found that most of the variation in annual growth rates could be related to environmental variability, and indeed, the model developed explained 80% of the variance in growth rate. However, the laboratory model for Brown Trout growth (Elliott, Hurley & Fryer 1995) strongly underestimated growth in some of the coldest rivers and a significant relationship was found between growth and temperature that explained ≈5% of the variance. This may indicate that thermal adaptation occurs in the coldest rivers.
Here, we investigated intraspecific variation in growth and feeding performance of Atlantic Salmon from five populations, originating from 59°N in south Norway to 70°N in north Norway, of which three are cold and two warm salmon rivers. We modelled experimental data for maximum growth, energy intake and gross growth efficiency relative to temperature, and evaluated the evidence for population-specific adaptive variation in these traits. We also compared the growth and feeding of groups of Atlantic Salmon parr exhibiting different growth rates (fast, moderate and slow growing individuals), thus presenting models for the lower and upper size modes frequently observed in Atlantic Salmon populations. Atlantic Salmon is known for its strong homing tendency (e.g. Hansen & Jonsson 1994), meaning that most fish return to their natal grounds for spawning. When straying occurs, most fish enter neighbouring rivers for spawning (Jonsson, Jonsson & Hansen 1991), and there is a high degree of reproductive segregation among populations (Ståhl 1987), which might promote population-specific adaptation to environmental conditions such as the temperature regimes of their home rivers.
Materials and methods
Experimental site and fish
All experiments were performed at the NINA Research Station, Ims, in southwestern Norway (59°N, 6°E). The fish used in the experiments were offspring of fish collected in the Rivers Alta (70°N, 23°E), Stryn (62°N, 7°E), Lone (60°N, 5°E), Suldal (59°N, 5°E) and Imsa (59°N, 6°E). The Rivers Stryn and Suldal are cold rivers with mean summer temperatures lower than 9 °C and annual mean temperatures of 6–6·5 °C (Table 1). The River Alta is somewhat warmer during summer with mean temperatures of 10·7 °C, but annual temperature is very low (≈4 °C) because of the very long winter. The Rivers Imsa and Lone are warmer with mean summer temperatures above 12 °C (Table 1). The Rivers Alta, Stryn and Suldal hold large salmon (mean sea age at maturity: 2·3–2·6 years) whereas one-sea-winter salmon dominate in the other two rivers (Table 1). Mean smolt age varies from less than 2 years in the warm River Imsa to nearly 4 years in the cold River Alta. Smolt size is similar among the rivers except in the River Imsa where the smolts are particularly large (Table 1).
Table 1. Environmental and life-history characteristics of the five populations of Norwegian Atlantic Salmon; the numbers of years studied (Nyears), the annual mean temperature (Tann°C), the annual amplitude of monthly mean temperatures (Tvar°C), the duration of the growth season (days), the average temperature during the growth season (Tseas°C), the maximum temperature (Tmax°C), the temperature sum during the growth season (Tsum°C), the average smolt age (years ± SD) and size (mm ± SD) and average sea age at maturity
For each laboratory-cohort, eggs were collected from a minimum of five females and fertilized with one male for each female. During incubation, the eggs were maintained at 2–7 °C. After hatching and during early feeding the temperature was 7–10 °C. After this period, and until acclimatization to experimental temperatures, the fish were maintained at ambient temperatures that varied to some extent among years but with a similar seasonal pattern.
Experiments were designed such that fish from four populations, five temperatures and two replicates could be run simultaneously in individual tanks. The 40 tanks were 45 × 45 cm2 and 60 cm deep, had a water flow of 3 l per min, water level of 30 cm and light intensity of ≈70 lux at the water surface during daytime. All experiments were performed at a constant day length (18 h). Experimental units were randomly distributed within each section to avoid systematic tank effects. Temperature was measured daily, and oxygen concentration was measured regularly, particularly in the experiments at higher temperatures. Oxygen saturation was always close to 100%.
Ten individually marked (Alcian blue in fins and adipose fin clipping) 0+ Atlantic Salmon were used in each tank. Before the experiments in February–March 1996, the fish were acclimatized to experimental conditions over a period of 40 days. This was done by gradually increasing the day length and temperature from the winter conditions at 10 h of light per day and ≈2 °C. During the last 7 days the fish were kept at the experimental temperatures. For the remaining experiments, the fish were acclimatized to experimental temperatures for 7 days because there was a less than 5 °C difference between holding and experimental temperatures. Each fish was weighed after 24 h of starvation at the beginning and end of each experiment. Fish that died during the experiments were replaced by similar sized fish to maintain densities, but the replacement fish were not included in the results. The fish were killed after each experiment, so no fish was used more than once so that observations were independent of each other. The fish were fed to satiation with CsCl-enriched granulated fish food (Cs concentration: 19 p.p.m. fresh mass and 23 p.p.m. dry mass) administered from automatic feeders. The Cs in the food was used to estimate food consumption by a tracer method (see below). The food was specially made by a research section of one of Norway’s fish food manufacturers (Felleskjøpet, Oslo, Norway).
One experiment was performed in 1996 and two in 1997 and 1998 with temperatures ranging from 5 to 24 °C (Table 2). Experiments were performed in two different years for all populations using offspring of different parents each year. Thus each population was represented by offspring of at least 10 females and 10 males. Each experiment lasted for approximately 3 weeks, and there was no break between experiments (apart from 1 week of acclimatization). Temperature ranges overlapped between experiments enabling a control for temporal changes in growth performance during the experimental period.
Table 2. Experimental year and period, preset temperatures (°C) and overall average mass (SD in parentheses) in experiments on the thermal performance of five populations of Norwegian Atlantic Salmon. At each temperature, two tanks with 10 fish in each tank were used. For modelling of growth and food consumption different experimental units were pooled to form 10 groups (see method section for further explanation)
16, 18, 20, 22 & 24
8, 10, 12, 14 & 16
11, 13, 15, 18 & 24
5, 7, 9, 10 & 16
16, 18, 20, 22 & 24
8, 10, 12, 14 & 16
11, 13, 15, 18 & 24
5, 7, 9, 10 & 16
5, 7, 9, 11 & 13
11, 13, 15, 18 & 24
5, 7, 9, 10 & 16
16, 18, 20, 22 & 24
8, 10, 12, 14 & 16
11, 13, 15, 18 & 24
5, 7, 9, 10 & 16
5, 7, 9, 11 & 13
16, 18, 20, 22 & 24
8, 10, 12, 14 & 16
The tracer method
Food consumption of Atlantic Salmon was estimated by a tracer method (Forseth et al. 1992), using stable caesium (133Cs). The estimation of food consumption is based upon estimating the intake of caesium from an observed change in caesium body burden with time. The rates of assimilation and elimination must be known, and food consumption is obtained by dividing the caesium intake by the concentration in food. Daily food ration (Di mg dry mass) was estimated by the Cs tracer method according to Forseth et al. (2001):
( eqn 1)
( eqn 2)
where Ii is daily caesium intake (µg), Qi and Qi+1 are caesium body burdens (µg) in fish at times i and i + 1, respectively, k1i and k2i are the rate constants of the fast and slow caesium elimination (day−1), respectively, Ci is the caesium concentration in the food (µg g−1 dry mass) and b is the assimilation of caesium from the food set at 0·79 (Forseth et al. 2001). A two-component caesium elimination model was used (Kevern 1966; Gallegos & Whicker 1971) and it was assumed that 18% of the caesium body burden was eliminated quickly and 82% slowly. Elimination rates were estimated from equations for Brown Trout (Ugedal et al. 1992) because elimination models are not available for Atlantic Salmon. Sensitivity analyses have shown that the estimated food consumption is insensitive to variation in elimination rate (Forseth et al. 2001) because the increase in Cs body burden is large during the experimental period.
Food consumption was estimated for each individual fish. Concentration of Cs in fish at the end of the experiments was determined for small groups of fish, grouped by their growth rate (see below) within experiments and temperatures (replicates pooled). For each temperature and experiment, the three main growth groups (fast, moderate and slow growing fish) were measured separately. Fish from each group were cut in small pieces, pooled, dried (48 h at 75 °C) and thoroughly homogenized before Cs was measured in a 0·4-g subsample by high-resolution ICP-MS (Finnigan, Bremen). All samples were weighed before and after drying to determine water content. Replicates were occasionally taken, and these showed that Cs values never deviated by more than 5%. Initial Cs body burdens in fish were set to a value close to zero (at 0·01 µg), representing background levels of caesium in fish. Variation in this value, even as large as one order of magnitude, did not influence results significantly. Caesium concentration in the granulated fish-food was measured in 10 random samples. The coefficient of variation in Cs concentration in these samples was approximately 5%. The water content of the food was also determined.
To be able to describe and model the range of individual growth rates occurring within the study populations, the fish were sorted into four groups according to their growth rates. As the experiments each year were performed on offspring of different parents, and the growth on some occasions differed between years, data from different years were pooled before groups were formed (Forseth et al. 2001). Firstly, following (Ostrovsky 1995), the standardized mass-specific growth rates (Ω percentage) were calculated as:
( eqn 3)
where M0 and Mt are the respective body masses (g) at the beginning and end of each experiment, b is the allometric mass exponent for the relation between growth rate and body mass, estimated at 0·31 for Atlantic Salmon by Elliott & Hurley (1997) and t is the experimental period (days). Secondly, replicates from the same or similar temperatures (within 1 °C) were pooled. Replicates did not differ significantly (Mann–Whitney U-test, P > 0·05). Third, individuals from each pooled set were sorted according to their standardized mass-specific growth rates and equal sized groups of fast, moderate and slow growing individuals were formed (the upper three quartiles). These quartile groups were used for the modelling of both growth and food consumption.
Growth and energy intake models
In accordance with Forseth et al. (2001), both growth and food consumption (measured in energy terms) were modelled by the ‘Ratkowsky model’, which is based on the four-parameter model of Ratkowsky et al. (1983) for the dependence of bacterial culture growth rate on temperature. The reparameterized model consists of two biological meaningful parameters: the lower (TL) and upper (TU) critical temperatures for growth and two parameters, d and g, that can be used to estimate the final two parameters of the Elliott et al. (1995) generic model for fish growth. The relationship between standardized mass-specific growth rates (Ω percentage) and temperature (T) is given by:
Ω = d(T − TL)(1 − e),( eqn 4)
and parameter values could be estimated using non-linear least squares. The third biological parameter, the optimal temperature for growth (TM), was estimated by solving:
ln[1 + g(TM − TL)] = −g(TM − TU).( eqn 5)
Finally, the height of the growth curve (c, the growth rate of a 1-g fish at the optimal temperature) was estimated by calculating equation 4 for T = TM. The Ratkowsky model is now expressed by the same four biologically interpretable parameters as the Elliott et al. (1995) model (TL, TM, TU and c).
The Elliott & Hurley (1998) consumption model for Brown Trout expressed the maximum daily energy intake (C kJ) in response to temperature (T °C) and mass (M g) in simplified form as:
C = ΦM0·766.( eqn 6)
An alternative model for Φ is the Ratkowsky model of equation 4, where the redefined parameters TL and TU are lower and upper critical temperatures for feeding, and d and g determine the temperature for maximal intake TM and the maximal value c. The energy value of the food pellets was 22·6 kJ g−1 dry mass.
Gross growth efficiency
The predicted gross growth efficiency, KG, for a fish of mass M0 was calculated from predicted growth and consumption in 1 day:
( eqn 7)
where M1 = (M0b + bΩ/100)1/b, Ω and Φ are given by the fitted Ratkowsky models (equations 4 and 6) for growth rate and consumption and J is the conversion factor of mass to energy for a fish of mass M0 (7 kJ g−1 fresh mass).
Evaluation of adaptive variation
Because we were unable to obtain identical temperatures each year it was not possible (owing to empty cells) to analyse the data in detail by use of analyses of variance. Overall differences in thermal performance, however, were explored by analysis of variance of the residuals from a common model (equation 4) for all populations. Furthermore, differences among populations in the biologically important traits (c, TL and TM) were explored by evaluating parameter values and their standard errors. Standard errors for the parameters in the growth and energy intake models were obtained using bootstrap (200 permutations) methods (Efron & Tibshirani 1993).
Within populations, the estimated model parameters differed in two aspects (Table 3). The height of the growth curves (c) declined and the breadth of the growth performance decreased with decreasing growth rate (among the three growth groups). The latter is caused by an increase in the lower (TL) and decrease in the upper temperature (TU) for growth. Judged from the standard errors of the parameter estimates most of these differences were statistically significant. The optimal temperature for growth remained more or less constant across growth groups.
Table 3. Estimated model parameters (± SE) for the thermal relationships of growth (standardized mass-specific growth rates, Ω, defined by equation 3) of Atlantic Salmon from five different Norwegian populations. The parameters in the model are the lower (TL) and upper (TU) critical temperatures for growth and the constants d and g that determine the optimal temperature for growth (TM) and the maximal value (c, the growth rate of a 1-g fish at the optimal temperature). Estimates are given for different groups of fish (fast-, moderate- and slow-growing individuals). Adjusted R2 is also tabulated
For all three groups there were significant effects of population (fast growing: F389,4 = 15,4, P << 0·01, moderate growing: F387,4 = 55·5, P << 0·01, slow growing: F386,4 = 102·1, P << 0·01) on the growth residuals (residuals from a common growth model). The optimal temperature for growth varied to some extent among populations (Table 3 and Fig. 1). It was highest for the River Alta (19·6–20 °C) and lowest for the River Suldal (16·3 °C) salmon. These two populations deviated significantly (judged from the standard errors) from the other three. The same trends hold for all three growth groups. There was no significant relationship between optimal temperature for growth and the temperature conditions in their rivers of origin (correlation analyses, variables tested are given in Table 1). The differences among populations in the lower temperature for growth were small.
Maximum growth rate at the optimal temperature, c, also varied among populations. The salmon from the River Stryn grew significantly (judged from the standard errors) faster than the others. This pattern was consistent for all three groups. Maximum growth rate was also high for the River Alta salmon and low for the River Suldal salmon compared with the two remaining populations. There was no consistent relationship between maximum growth rate and optimal temperature for growth. The growth rates for the different populations were similar from 7 to 15 °C, the common temperature range during the growth season, except the River Stryn salmon, which grew faster than the others between 10 and 20 °C.
The temperatures for maximum energy intake were estimated at ≈19–21 °C for all populations (Table 4 and Fig. 2), slightly higher than the optimal temperatures for growth (Table 3). The lower temperatures for consumption were on average nearly 2 °C lower than the lower temperatures for growth, whereas the upper temperatures for consumption were on average approximately 1·5 °C higher than the corresponding limits for growth. The energy intake at the temperature for maximum consumption (c) for moderate and slow growing fish was on average approximately 70 and 50%, respectively, of that for fast growing fish. However, this difference varied from 60 and 30% for the River Suldal salmon to 80 and 60% for River Stryn salmon. Apart from this decline, the lower temperature for food consumption was the only other model parameter that changed consistently across growth-groups.
Table 4. Estimated model parameters (± SE) for the thermal relationships of energy intake (standardized daily energy intake, Φ, defined by equation 6) of Atlantic Salmon from five different Norwegian populations. The parameters in the model are the lower (TL) and upper (TU) critical temperatures for energy intake and d and g that determine the temperature for maximum energy intake (TM) and the maximal value (c) at TM. Estimates are given for different groups of fish (fast, moderate and slow growing individuals). Adjusted R2 is also tabulated
For all three groups there were significant effects of population (fast growing: F389,4 = 18,4, P << 0·01, moderate growing: F388,4 = 43·5, P << 0·01, slow growing: F386,4 = 55·5, P << 0·01) on the energy intake residuals (residuals from a common energy intake model). Thus the growth and consumption curves were functionally similar, and the among-population variation in consumption paralleled that of the growth rate, particularly in that salmon from the Rivers Stryn and Alta performed better than the others. Accordingly, the growth rate correlated significantly with the energy intake in all populations studied (P << 0·01). The fits, however, varied (R2 0·45–0·80) and was relatively poor for the River Imsa population (R2 0·45).
The modelling of growth efficiencies showed that efficiency maxima were very high, 42–58% (Fig. 3, estimated for moderately fast growing fish). Efficiencies were highest for salmon from the cold rivers in south Norway, Suldal and Stryn, which both live in hostile environments with low individual growth rates in their natal environment. It was lowest for fish from the cold River Alta from north Norway and from the warmer rivers from south Norway, Lone and Imsa, of which the latter two may be classified as benign environments given the high growth rates of the parr in those rivers and the long growth season. The efficiency reached a maximum at 12 °C for the River Suldal salmon, 18 °C for the River Alta salmon and approximately 15 °C for the other populations studied. There was no distinct peak in the growth efficiency but the temperature for maximum efficiency was lower (approximately 1·5–4·5 °C) than the corresponding temperature for maximum growth and food consumption.
There was no clear pattern in growth performance either in relation to latitude or temperature regime in the river of origin, and we found little support for either of the two hypotheses on thermal adaptation (Levinton 1983; Conover & Schultz 1995). There were significant differences among populations in both growth and consumption but these differences appeared unrelated to thermal conditions in the rivers. There was no significant relationship between optimum temperatures (TM) for growth and the prevailing temperatures during the growth season in their rivers of origin. Moreover, the lower and upper temperature limits (TL and TU) and growth rate at maximum growth, c, did not vary among populations as predicted by any of the hypotheses on thermal adaptation.
The only pattern that may accord with the counter-gradient variation hypothesis was the superior performance of fish from the River Stryn. In this cold, glacier-fed river, rapid growth may be important for survival and smolting (physiological, morphological and behavioural changes preadapting the fish for sea life), and thus the fitness of the fish. However, salmon from the River Suldal, which also is cold and glacier fed, did not show any similar enhanced growth, and the fish from the river with the shortest growth season, the River Alta, did not grow particularly fast. Moreover, the growth pattern in the River Alta was very similar to that of the salmon from the River Lone where the growth season is much longer (nearly twice as long) and summer temperatures higher (see Table 1). Moreover, the Norwegian salmon in the present study had similar growth performance to that found in two British populations (Elliott & Hurley 1997), except that TM was lower for the salmon from northern England, the opposite of the prediction from the hypothesis of adaptation to a local optimum (Levinton 1983). All in all, this gives support to Elliott’s (1994) hypothesis of little if any population-specific adaptation in growth performance to local climatic conditions in Atlantic Salmon.
Is the lack of population-specific thermal adaptation in Norwegian salmon a result of genetic influence of escaped farmed salmon entering most populations? This possibility cannot be ruled out, but the relatively large and significant variation in thermal performance among the populations explored indeed indicates that the populations are not swamped genetically by escapees. If all populations were strongly influenced by the gene pool of the common Norwegian farmed salmon we would expect little if any variation in performance among populations. Furthermore, recent studies on life-history variation among Norwegian salmon (Jonsson, N. et al. 1991, 1998) indicate that population-specific differences are still maintained, which should not be the case if our results were due to a strong genetic influence from farmed salmon. We conclude that there is variation in growth and feeding potential among salmon populations and this variation may be adaptive, but we found no pattern related to thermal conditions in their respective natal rivers.
High heritability has been estimated for growth rates in Atlantic Salmon (Gunnes & Gjedrem 1978) and is the basis for the rapid selection of production traits in salmon farming (Gjerde 1986). Moreover, inherited differences in growth among populations of ectothermal animals are well established (Berven & Gill 1983; Niewiarowski & Rosenburg 1993). Why then, does not Atlantic Salmon exhibit any adaptation in growth performance to the local climate of the home river? It could be argued that because these fish spend years of their life at sea, and the European populations presumably share feeding habitat during the marine phase of the cycle (Friedland et al. 2000), low interpopulation variability is to be expected. However, the present models and experimental data were based on freshwater growth of immature fish. During this phase, thermal conditions vary widely among rivers, and this could have generated differences in thermal performance. Indeed, Nicieza et al. (1994) gave evidence of counter-gradient variation in digestive performance of Atlantic Salmon from Spain and Scotland. It is also surprising that the temperature for maximum growth (TM) in all the populations is much higher than the temperature the fishes normally experience both in freshwater and at sea.
Our experiments and modelling provided one interesting difference among populations. Among the rivers in south Norway, the fish from the rivers with poor growth conditions (Stryn and Suldal) showed higher growth efficiency than those from the richer rivers, Lone and Imsa. The north Norwegian River Alta salmon stands out as being the one with the lowest growth efficiency. One important difference between River Alta and the cold rivers in south Norway (Stryn and Suldal) is that the winter is nearly 3 months longer in the north. Under such conditions, storage of lipid energy at the expense of growth in structural tissues may be particularly important for surviving the long winter. This implies that efficiency may indeed be high in terms of energy. This distinction cannot be made in the present study because we used a common conversion factor to estimate the energy of growth. Thus, differences among populations in growth efficiency may result from a combined effect of temperature, seasonal effects and feeding conditions rather than temperature alone. Natural selection may favour higher growth efficiency under poor feeding conditions such as those in the Suldal and Stryn, where the fish need about 50% longer to smolting than the fish in the warmest rivers in this study, Imsa and Lone.
Wootton (1998) maintained that there is a trade-off between growth rate and growth efficiency. Possibly, the salmon maximizes growth rate when feeding opportunities are good, and growth efficiency under hostile feeding conditions. We know of no other study from fish showing a similar possible adaptation to variable feeding conditions, and it is at least partly contrary to the conclusion of Schultz, Reynolds & Conover (1996) of increasing growth rate with latitude, as they found in Killifish (Fundulus heteroclitus). In an analysis of the swimming speed of young Bleak (Alburnus alburnus (L.)), Ware (1975) suggested that the fish swam at a speed maximizing their growth rate rather than their growth efficiency. This is compatible with optimal foraging theory, which assumes that fish forage in a way that maximizes their net rate of energy intake (e.g. Hart 1986). Under extremely poor feeding opportunities, however, maximizing the gross energy conversion or growth efficiency may be the optimal choice for the fish.
The present application of the Ratkowsky et al. (1983) model shows that the growth of Atlantic Salmon varies to some extent among populations, but the variation appears unrelated to thermal conditions. The growth models for Norwegian salmon developed here are similar to those from two British salmon populations given by Elliott & Hurley (1997). Thus, if models are selected based on some knowledge of growth condition in the rivers (fast or slow growing salmon) the developed models are appropriate for predicting growth and consumption of Atlantic Salmon parr coming from different environments. The models developed for moderately fast growing fish are probably suitable for predicting average growth in the populations (Forseth et al. 2001), whereas the models for fast and slow growing fish are suitable for predictions of the upper and lower size modes, respectively.
We thank the staff at NINA Research Station for valuable assistance during the experiments, and June Breistein for technical assistance in experiments, laboratory treatment of the material and calculations. Peder Fiske and Margaret Hurley provided bootstrap estimates of standard errors on model parameters and we are thankful for their assistance. This study was supported financially by the European Commission under the FAIR-programme (contract no. CT95-0009) and the Norwegian Research Council under the Biological Diversity – Dynamics, Threats and Management programme and the Climate and Ozone programme.