Density-dependent growth in brown trout: effects of introducing wild and hatchery fish

Authors


T. Bohlin, Animal Ecology, Department of Zoology, Göteborg University, Box 463, SE-405 30 Göteborg, Sweden. Fax: + 46 31 416729; E-mail: torgny.bohlin@zool.gu.se

Summary

  • 1Although it is not clear to what extent density dependence acts on the survival, emigration or growth of organisms, experiments testing alternative explanations are rare. A field experiment on 1-year-old brown trout ( Salmo trutta L.) was undertaken to address the following questions: are the mortality, movement and growth of wild stream-living trout affected by population density? If so, are the density-dependent effects of released hatchery trout different from the effects of wild fish?
  • 2In each of two small streams, two replicate treatment blocks were used, each with four treatments assigned to stream sections 50–70 m in length: (1) control, no fish was introduced and population density was kept at its original level. (2) Trout biomass was doubled by introducing additional wild fish. (3) Trout biomass was doubled by introducing additional hatchery fish. (4) Hatchery trout were introduced, but biomass was kept at its original level by the removal of some resident wild fish.
  • 3We found no treatment effects on the recapture rates of resident trout, which suggests that survival was not strongly affected by competition. They were also remarkably stationary, regardless of treatment. However, trout growth rate was reduced to the same extent in both treatments with increased density, suggesting that growth was negatively density-dependent, and that the density-dependent effects of hatchery trout and introduced wild fish were similar.
  • 4Wild resident fish grew faster than introduced wild trout, which in turn grew faster than hatchery trout. Hatchery fish and introduced wild fish moved more than wild resident fish.
  • 5The results show that population density affected growth in trout parr. We conclude that competition is not limited to the underyearlings, as has previously been suggested, and that density-dependent growth is the main density-dependent response in yearling trout. Furthermore, this effect was the same for wild and hatchery-reared competitors, suggesting that stocking of hatchery fish may affect natural populations negatively through density dependence.

Introduction

One of the fundamentals of biology is that no population can increase without limit and, consequently, that population growth generally is believed to be negatively density-dependent. Density dependence is caused by a decrease in age-specific fecundity, survival and/or increased migration with increasing population density. In fish and other organisms with indeterminate growth, density dependence may act on individual growth, which primarily would affect fecundity. Based on studies in one English stream, Elliott (1994) concluded that mortality and emigration in small fry are the main density-dependent factors regulating the population sizes of brown trout (Salmo trutta L.) in streams, and that average individual growth is less affected by population density (Elliott 1989a,b). However, other studies have shown a negative relationship between abundance and average body size (Grant & Kramer 1990; Bohlin et al. 1994) suggesting that growth may be reduced at high densities. Jenkins et al. (1999) tested this idea by manipulating brown trout densities in several streams over a number of years, and found that the growth of brown trout was negatively related to density in most years, whereas movement and mortality were unrelated to stocking density. Further experimental studies are thus required to evaluate the relations between density and growth in this species.

Knowledge of density-dependent mechanisms is also required to predict the effects of hatchery-released or escaped farm fish on wild populations (Jonsson & Fleming 1993; Jonsson 1997; Thorpe 1998). Hatchery-reared fish have often lower growth, fecundity and survival in the wild than resident wild fish (Skaala, Jorstad & Borgstrom 1996; Jonsson 1997; Olla, Davis & Ryer 1998; Deverill, Adams & Bean 1999). However, released farm salmon may reproduce successfully in the wild, which may reduce the productivity or persistence of wild populations through ecological and genetic mechanisms (McGinnity et al. 1997; Fleming et al. 2000). In a field experiment, Weiss & Schmutz (1999) released domesticated brown trout in one crystalline and one limestone stream and found no consistent effects on the mortality, growth and movements of resident fish wild trout in the two streams. Thus, it is unclear to what extent, and under what conditions, the release of domesticated fish will induce density-dependent effects on survival, movements and individual growth in wild populations.

In this study, we manipulated brown trout densities in two streams to address the following questions: are the mortality, growth and movement rates of wild stream-living trout affected by population density? If so, are the density-dependent effects of released hatchery trout different from the effects of wild fish?

Methods

study area and populations

The experiment was conducted between May and August 2000 in sections of two small coastal soft-water streams in SW Sweden, Jörlanda and Norum (58°N). Jörlanda has a catchment area of 39 km2, and a mean width (depth) in the experimental areas ranging from 2·6 m (0·21 m) (upper part, 8 km from the outlet) to 3·7 m (0·27 m) (lower part, 4 km from the outlet). Conductivity ranges from 7 mSm−1 (upper part) to 10 mSm−1 (lower part). Norum has a catchment area of 19 km2, with a mean width (depth) in the experiment area (3 km from the outlet) of 3·0 m (0·28 m), and a conductivity of 14 mSm−1. Both streams are surrounded and shaded by mainly deciduous forests (mainly alder), grading to arable and pasture lands that dominate the lower part of the catchment. In the experimental sections, the substrates were mainly gravel-stone with little macrophytic vegetation. Both streams are characterized by large changes in water flow, usually with maximum flow in winter, and low-water mainly during summer.

The study populations were stages of sea-run brown trout, which is the only fish species in Norum, and the most abundant fish (> 95% by biomass) in Jörlanda. The trout in both streams generally spend their first 2 years in the stream before migrating to the sea (Bohlin, Dellefors & Faremo 1994), but varying proportions of these populations, mainly males, remain in the streams throughout their lives (Dellefors & Faremo 1988). In the study area, the average population densities in autumn for all age-classes fluctuate around approximately 1·3 m−2 in both streams.

To avoid confusion, we use the terms hatchery fish for the offspring of wild parents that are reared in a hatchery, and domesticated fish for trout reared in the hatchery over multiple generations. Further, the wild trout in this study were of two kinds. Those that were tagged and released at the site of capture will be called wild resident trout, whereas the wild trout that were moved from one site to another are referred to as wild introduced trout.

In this study, we released 1-year-old hatchery trout, offspring from wild parents collected from the study streams, and reared under standard conditions in Båkab Fish Farm (Laholm, Sweden). The length distributions of the released hatchery populations are given in Fig. 1.

Figure 1.

Length distributions of experimental fish of wild and hatchery-reared brown trout in Jörlanda and Norum.

experimental design

Experimental unit and target fish

The experimental unit was a stream section with a length of 50 m (Jörlanda) and 70 m (Norum), subdivided into 10 m subsections by labelled markers on the bank. The target fish were the wild resident trout individuals in the units. In May 22–26 2000 (time 0), these were captured by two electrofishing runs (LUGAB 1000, straight DC, 200–400 V), anaesthetized (2-phenoxyethanol, 0·1 mL/l), tagged with passive integrated transponders (Trovan, ID 100), measured (fork length to the nearest mm), weighed (to the nearest 0·1 g) and released into the 10-m subsection where captured. In total, the resident populations consisted of 650 tagged individuals (Table 1). The initial length distributions of resident fish are given in Fig. 1.

Table 1.  Number of resident brown trout (target fish) tagged and number of introduced wild and hatchery trout in the experimental units. HL, hatchery fish at low density; HH, hatchery fish at high density; W, wild fish at high density; and C, control group. Blocks are upstream (U) and downstream (D) in relation to each other. Total initial biomass is given g m −2   ×  100
StreamTreatmentBlockNumber of fish taggedTotal initial biomass
Target fishWild intr.Hatchery fishTotal number
JörlandaHLD27    023  50  590
JörlandaHLU24    026  50  620
NorumHLD26    013  39  494
NorumHLU22    013  35  307
JörlandaHHD47    039  86  939
JörlandaHHU48    031  791330
NorumHHD33    042  75  778
NorumHHU60    037  97  813
JörlandaWD46  50  0  961070
JörlandaWU36  55  0  911070
NorumWD71105  0176  457
NorumWU56  75  0131  501
JörlandaCD48    0  0  48  670
JörlandaCU35    0  0  35  371
NorumCD23    0  0  23  255
NorumCU49    0  0  49  269

Treatments and experimental fish

The responses of the resident fish to the following four treatments were studied:

  • 1C (control): no additional fish were introduced so that population density was kept at its original level.
  • 2HH (hatchery high): hatchery trout were introduced so that total trout biomass in the unit was approximately doubled.
  • 3HL (hatchery low): hatchery trout were introduced, but the total trout biomass in the unit was kept at its original level by removal of some of the resident trout.
  • 4W (wild): wild trout, with a size distribution similar to that of the target fish, were introduced so that total trout biomass in the unit was approximately doubled.

The introduced wild trout were obtained partly from the HL treatment and partly from areas downstream (300–1000 m) of the experimental sections. Each introduced fish was measured and weighed, tagged with a passive integrated transponder (Trovan, ID 100), and released evenly into the experimental units while noting the subsection label of release. The number of fish in each treatment is given in Table 1.

Blocking and physical arrangements

The four treatments were arranged in blocks. Within each block, the four experimental units were selected from what we judged as good habitat for juvenile trout, and in such a way that the units were as environmentally similar as possible. Within blocks, treatments were assigned at random. The experimental units within the blocks were separated by 30–300 m depending on environmental variation within blocks. We used two blocks in each of the two streams, resulting in a total of 16 treatment units. In Jörlanda, the two blocks were 3·5 km apart, whereas in Norum they were adjacent. Each experimental unit was screened with nets (mesh 8 mm) during the first week of the experiment. However, because of heavy rainfall the day after the introduction of the experimental fish, the net enclosures did not prevent trout from straying. The nets were removed after 1 week.

Recapture

Two electrofishing recaptures were conducted in each stream, the first 26–29 days after release (recapture 1, June 17–21 in Jörlanda, and June 26–27 in Norum), and the second 84–91 days after release (recapture 2, August 14–16 in Jörlanda and August 21–22 in Norum). Two electrofishing runs were conducted in and between the treatment units in each block, including a 50-m stream section immediately upstream and downstream of each block. In addition, one electrofishing run was conducted in each stream, from 500 m downstream of the downstream block to 500 m upstream of the upstream block.

statistical treatment

Response variables

The response variables were proportion recaptured, absolute distance moved, relative change in condition, and growth in weight of the target fish (wild resident fish). Estimates were from three periods; period 0–1 (from tagging to first recapture, 26–29 days), period 0–2 (from tagging to second recapture, 84–91 days), and period 1–2 (from first to second recapture, 58–63 days). For recapture rate, however, we did not separate trout recaptured at times 1 and 2 or both; an individual was either recaptured or not recaptured. The relative condition of each fish captured was estimated as its residual from the log(weight)-log(length) regression over all treatments, separately for each time (times 0, 1 and 2) and stream, from which the change in condition, Cdiff, over the periods was calculated. The specific growth rate (Gw) in weight for recaptured individuals was calculated as

Gw   =  100(ln Wt   −  ln W0 ) t −1

where W0 is initial body size and Wt is body size t days later.

Statistical models

Responses on growth, movement and change in condition were analysed using ordinary analyses of covariance, whereas recapture rates were analysed using a logistic model. Since the treatments can be arranged as a 2 × 2 factorial with the fixed factors Density (levels Low and High) and Origin of the introduced competitors (levels Hatchery and Wild), we used the following general model:

Response variable = Stream + Block (Stream) + Origin + Density + Origin  ×  Density + Initial Length (model 1)

Stream is a random two-level class variable, Block a random class variable nested in Stream, and Initial Length a continuous variable (covariate). In the treatment level Hatchery in the factor Origin, it can be noted that although only hatchery trout were added, both wild and hatchery trout were present in the units.

We also used the experiment to compare the performance (growth, recapture, movement, condition change) between the categories wild resident (W) and wild introduced (I) individuals, and between the categories wild resident (W) and hatchery trout (H). For the W–I comparison, we used only fish from treatment W to avoid Density as a confounding factor, and applied model 2:

Response variable = Stream + Block (Stream) + Category + Initial length

For the W–H comparison we used, for the same reason as above, only trout from treatments HH and HL, resulting in model 3:

Response variable = Stream + Block (Stream) + Category + Density + Initial length

where Density is a class variable with levels High (treatment HH) and Low (treatment HL).

Results

sampling efficiency of electrofishing

Using the maximum likelihood solution for the two-run removal estimator (Seber & Le Cren 1967) the estimated catch probability of the target fish in one removal was 0·624 (SE = 0·0393), suggesting that about 86% of the population was captured in the two-run sampling.

treatment effects on the resident fish

Recapture

Recapture rates were affected by Stream and Block, but not by Initial length, Density, Origin or Density × Origin (Table 2). The overall recapture rates were 61·4% in Jörlanda and 45·5% in Norum.

Table 2.  Treatment effects on wild resident brown trout (target fish) individuals. Test results ( P , F and MSE) of Model 1. Degrees of freedom (d.f.) given within brackets. Growth is the specific growth in weight, C diff is the change in condition (weight scaled to length) and Mov the absolute distance the fish had moved. Numbers 0–1, 0–2 and 1–2 indicate the periods (0–1 from tagging to first recapture, 0–2 from tagging to second recapture, and 1–2 from first to second recapture). For Density effects on growth, see Fig. 3
Response variableModel P , R2 , d.f., MSE Background variablesOrigin of competitors (1 d.f.) P, F MSEDensity of competitors (d.f.) P, F MSEOrigin × density (1 d.f.) P, F MSE
Stream (1 d.f.) P , F MSE Block (2 d.f.) P , F MSE Initial length (1 d.f.) P, F MSE
  • a

    Not included in the model.

  • b

    Logistic regression.

Growth 0–1< 0·0001, 0·400·8254, 0·050·0022, 6·31< 0·0001, 114·80·1360, 2·240·0411, 3·29 High < Low0·2740, 1·20
(Error d.f. = 224)7, 0·08640·004210·54529·9170·19340·28410·1039
Growth 0–2< 0·0001, 0·380·0067, 7·500·1468, 1·94< 0·0001, 117·30·8519, 0·03< 0·0001, 17·43 High < Low0·5002, 0·46
(Error d.f. = 219)7, 0·02930·22000·05673·4380·00100·51100·0113
Growth 1–2< 0·0001, 0·37< 0·0001, 21·100·0261, 3·77< 0·0001, 32·150·6747, 0·180·0013, 10·87 High < Low0·8285, 0·050
(Error d.f. = 109)7, 0·03060·64480·11520·98240·00540·33200·0014
Cdiff 0–1< 0·0001, 0·31< 0·0001, 66·81< 0·0001, 13·20a0·0537, 5·190·7709, 0·090·1632, 1·96
(Error d.f. = 225)6, 0·000960·06410·0127 0·00500·00000·0019
Cdiff 0–20·0002, 0·110·0002, 14·170·0037, 5·74a0·2386, 1·400·6506, 0·210·9331, 0·01
(Error d.f. = 220)6, 0·001750·02480·0100 0·00240·00040·0000
Cdiff 1–20·0015, 0·180·0493, 3·950·0069, 5·21a0·8970, 0·020·0305, 4·80 High < Low0·4665, 0·53
(Error d.f. = 110)6, 0·000900·00360·0047 0·00000·00430·0005
Recaptureb (d.f. = 448)< 0·0001< 0·00010·03310·53970·64240·73630·8610
   χ 2   =  14·79 χ 2   =  6·82 χ 2   =  0·38 χ 2   =  0·22 χ 2   =  0·11 χ 2   =  0·03

Movements

For trout movements, none of the models were significant (period 0–1: P= 0·684, d.f. = 7,224, F= 0·02; period 0–2: P= 0·212, d.f. = 7,219, F= 0·04; period 1–2: P= 0·377, d.f. = 7,109, F= 0·02). Large proportions of resident fish were recaptured in the subsection where last recorded (Fig. 2). During period 0–1, 85% of the recaptures were made within 20 m of release points, and 57% within 10 m (that is, in the subsection where the fish was last recorded). The corresponding figures for periods 1–2 and 0–2 were strikingly similar, 85% and 80% within 20 m, and 57% and 56% within 10 m, respectively.

Figure 2.

Distance moved by wild resident brown trout (target fish) from tagging to second recapture. Positive numbers: upstream direction.

Change in condition

The change in condition was related to Stream and Block (Table 2). In periods 1–2, the factor Density was significant; the loss in condition was larger in the high-density than in the low-density treatments (Table 2). Origin or Density × Origin had no significant effect (Table 2).

Growth

Growth during all three periods was affected by Density. Origin and Density × Origin were, however, not significant (Table 2). Individual growth in the two treatments where biomass was increased (HH and W) was thus reduced to a similar extent, compared to the C and the HL treatments (Fig. 3). The lack of an interaction effect indicates that introduced fish affected resident trout to a similar extent regardless of the origin of the former. As expected (Elliott & Hurley 1999), growth declined with initial body size.

Figure 3.

Effects of density (factor Density in model 1) on the growth (specific growth, day −1   ×  100) of wild resident trout (the target fish), given as least square means with standard error bars. *** P  < 0·0001, *0·01 < P  < 0·05.

performance of wild and hatchery fish

Using the maximum likelihood solution for the two-run removal estimator (Seber & Le Cren 1967), the estimated one-catch probabilities of the resident fish, the introduced wild trout, and the hatchery trout were 0·624 (SE = 0·0393), 0·698 (SE = 0·0582) and 0·611 (SE = 0·0800), respectively. This suggests that two runs would capture 84–91% of the fish. We found no significant difference between the three categories in the distribution of the two catches (χ2 = 0·0509, d.f. = 2), indicating that the catch probability was similar among the categories. In the comparison between wild resident and hatchery trout (W–H), we found significant differences in growth, change in condition and movements (Table 3). Initially, resident trout grew faster, and over the season they lost less in condition and moved less than the hatchery fish (Fig. 4). We did not find any effect of Density (HL vs. HH) on growth in this case (Table 3). In the comparison between the wild resident trout and the wild introduced trout, the pattern was similar as between wild and hatchery trout (Table 4, Fig. 4) but the differences were generally smaller, and we found no significant differences in change in condition between the trout categories (period 0–1: P= 0·141, d.f. = 5,133, F= 1·69; period 0–2: P= 0·166, d.f. = 5,155, F= 1·59; period 1–2: P= 0·685, d.f. = 5,70, F= 0·62). Nor did we find any significant differences in recapture rates between categories in any of the tests, although recapture rates tended to be higher for wild resident trout than for hatchery trout (W: 53%, H: 49%) and higher for resident than introduced wild trout (W: 53%, I: 46%) (Tables 3 and 4).

Table 3.  Performance of the categories wild resident trout and hatchery trout in the HH and HL treatments. Test results of Model 3. Response variables are explained in Table 2. For differences between the Categories, see Fig. 4
Response variableModel P , R2 , d.f., MSE Background variablesCategory (wild, hatchery) (1 d.f.) P , F MSE
Stream (1 d.f.) P , F MSE Block (1 d.f.) P , F MSE Density1 (1 d.f.) P , F MSE Initial length (1 d.f.) P , F MSE
  • a

    Not included in the model.

  • b

    Logistic regression.

Growth 0–1< 0·0001, 0·720·0104, 6·700·0006, 7·700·8738, 0·03< 0·0001, 116·6< 0·0001 , 144·2
(Error d.f. = 189)6, 0·0700·47110·54140·00188·19910·14, Wild > Hatch
Growth 0–2< 0·0001, 0·560·0449, 4·100·1008, 2·330·0881, 2·95< 0·0001, 72·770·0024 , 9·59
(Error d.f. = 138)6, 0·02890·11850·06750·08532·1050·2774, Wild > Hatch
Growth 1–20·0003, 0·280·0236, 5·340·1415, 2·010·1140, 2·56< 0·0001, 19·700·4871, 0·49
(Error d.f. = 75)6, 0·033540·18880·07100·09040·69640·0172
Cdiff 0–1< 0·0001, 0·72< 0·0001, 85·5< 0·0001, 24·60·7548, 0·10a< 0·0001, 344·8
(Error d.f. = 190)5, 0·0000510·00440·001270·000005 0·0177, Wild > Hatch
Cdiff 0–2< 0·0001, 0·530·0043, 8·440·0007, 7·740·7640, 0·09a< 0·0001 , 131·0
(Error d.f. = 139)5, 0·000110·000950·0008670·00001 0·0148, Wild > Hatch
Cdiff 1–20·0066, 0·180·6997, 0·150·0090, 5·020·3578, 0·86a0·0371 , 4·50
(Error d.f. = 76)5, 0·0000620·0000090·000310·000053 0·0003, Wild > Hatch
Mov 0–10·0033, 0·100·6454, 0·210·2291, 1·490·7923, 0·070·4657, 0·530·0007 , 11·99
(Error d.f. = 188)6, 110·323·44163·97·6758·941323, Hatch > Wild
Mov 0–2< 0·0001, 0·200·0564, 3·700·0141, 4·390·0401, 4·290·5705, 0·320·0006 , 12·49
(Error d.f. = 138)6, 50·4186·6221·6216·516·30629·6, Hatch > Wild
Mov 1–20·0178, 0·180·0844, 3·060·0715, 2·730·0641, 3·530·2164, 1·550·1733, 1·88
(Error d.f. = 75)6, 25·477·5369·3089·539·4047·6
Recaptureb< 0·00010·087210·080·98920·08720·0713
   χ 2   =  2·93 χ 2   =  2·93 χ 2   =  0·00 χ 2   =  2·93 χ 2   =  3·25
Figure 4.

Comparisons of the performance of wild resident trout and stocked hatchery trout (left panel, data from the HH and HL treatments, model 2) and wild resident trout and wild introduced trout (right panel, data from the W treatment, model 3), given as least square means with standard error bars. *** P < 0·0001; **0·0001 <  P  < 0·01; *0·01 <  P  < 0·05, NS P > 0·05. Change in condition and distance moved are given only for the whole period (period 0–2).

Table 4.  Performance of the categories resident trout (target fish) (Res.) and introduced wild trout (Intr.) in the W treatment. Test results of Model 2. Response variables are explained in Table 2. For differences between the Categories, see Fig. 4
Response variableModel P , R2 , d.f., MSE Background variablesCategory (1 d.f.) P, F MSE
Stream (1 d.f.) P , F MSE Block (2 d.f.) P , F MSE Initial length (1 d.f.) P, F MSE
  • Not given as the model was not significant.

  • Logistic regression.

Growth 0–1< 0·0001, 0·420·8807, 0·02< 0·0001, 13·40< 0·0001, 157·20·0036 , 8·59
(Error d.f. = 304)5, 0·08910·00201·1914·000·765, Res > Intr
Growth 0–2< 0·0001, 0·350·0002, 14·010·0342, 3·41< 0·0001, 129·20·0064 , 7·53
(Error d.f. = 298)5, 0·03190·44700·10894·120·2401, Res > Intr
Growth 1–2< 0·0001, 0·25< 0·0001, 26·410·0510, 3·03< 0·0001, 28·660·7702, 0·09
(Error d.f. = 151)5, 0·03620·95540·10981·0370·0031
Mov 0–1< 0·0001, 0·120·5962, 0·280·1126, 2·200·0124, 6·190·0001, 27·69
(Error d.f. = 304)5, 88·324·8194·2546·52444·2, Res < Intr
Mov 0–2< 0·0001, 0·290·0084, 7·080·0010, 7·070·8252, 0·050·0001, 77·28
(Error d.f. = 298)5, 59·8423·7423·12·924623·9, Res < Intr
Mov 1–20·4654, 0·03
(Error d.f. = 151)5, 108·0    
Recapture< 0·00010·12520·00080·38040·0715
   χ 2   =  2·35 χ 2   =  14·32 χ 2   =  0·77 χ 2   =  3·25

Discussion

The first published experimental tests of density-dependent performance in stream-living salmonid fish, conducted on fry during their first summer, showed that survival and individual growth declined with initial stocking density (Frazer (1969), coho salmon (Oncorhynchus kisutch (Walbaum)) and steelhead trout (Oncorhynchus mykiss (Walbaum); Le Cren 1973, brown trout). There are few similar experiments on older stages. Jenkins et al. (1999), however, manipulated the density of brown trout and found effects on growth, but not on survival and movements, in both yearlings and underyearlings. In addition, the negative relationships found between body mass and density in salmonids (Grant & Kramer 1990; Bohlin, Dellefors & Faremo 1994) are indications of density dependence, although not necessarily of density-dependent growth. Observational time-series data on the young stages of migratory brown trout (Elliott 1994) showed strong density-dependent mortality during the first summer but no effects on average individual growth rates. However, Jenkins et al. (1999) found little density dependence in growth at single locations when examining their relationships over time. Elliott's samples were from only one location, which may explain why he did not find density-dependent growth.

In the present study, density affected growth in trout parr (underyearlings and older). Our conclusion is thereby consistent with that of Jenkins et al. (1999); a lower population density, due to higher initial mortality and/or lower egg deposition rates, leads to compensatory individual growth in trout parr. This suggests that competition is not limited to young fry, as Elliott (1994) suggested, and that density effects on growth in trout older than one summer may regulate population sizes. Further, density effects on resident target fish were similar regardless of whether the competitors were wild or hatchery-reared. Stocking of hatchery fish may thus affect wild individuals negatively, even if the stocked individuals are older than the age-class that Elliott suggested to be the regulated (bottle-neck) stage.

In contrast to many studies conducted on younger (smaller) stages of salmonids, but in agreement with Jenkins et al. (1999), we found no indication of density effects on survival. This might be expected because small individuals have higher metabolic rates per unit body mass and less nutrient reserves than large individuals, and may thereby suffer higher mortality during periods of severe competition (Sibly & Calow 1986; Roff 1992).

The movement distances of resident trout were independent of both density and body size (Table 2) and strikingly limited even after three months. The spatial distribution of recaptures (recapture effort covering at least ± 500 m (Fig. 2) excludes that movements were strongly underestimated because of a limited sampling distance. More than half of the recaptures were made within 10 m of the point of release, and the great majority within 20 m, despite the disturbance that may have been caused by repeated electrofishing (Nordwall 1999). Thus, even in the high-density treatments, individuals tended to remain at their original sites, despite increased competition. This would suggest that the fitness costs of leaving a site and searching for a better alternative are high, a view supported by the observation that wild individuals that were transferred into a new habitat grew initially slower than the resident trout (Fig. 4). A possible explanation is that prior residents often have a competitive advantage over intruders (Bradbury & Vehrencamp 1998; Deverill et al. 1999; Johnsson, Nöbbelin & Bohlin 1999; Johnsson, Carlsson & Sundström 2000). In contrast, in a stocking experiment with domesticated trout in a limestone stream, the movements of wild resident increased with increased stocking density (Weiss & Schmutz 1999), indicating that domesticated trout can displace wild ones. On the other hand, in the same study but in a crystalline stream, more than 20% of the wild fish moved in all treatments, with no relationship to stocking density. In both of these experiments, the trout (introduced and resident) were larger than in the present study (2–3 years old). Recent evidence suggest that large trout may move up to a hundred metres over a diel cycle (Young 1999), and considerably further over a season (Gowan et al. 1994; Gowan & Fausch 1996). Although absolute distances were not given, Ombredane, Bagliniere & Marchand (1998) reported that the distance of downstream movement in juvenile brown trout was positively correlated with body size. The results of the present study are, however, evidence for the traditional view that brown trout parr in small streams show restricted movements (Fig. 2), regardless of body size (Table 2), at least during summer.

Hatchery trout grew slower and lost more in condition than wild fish (Table 3, Fig. 4). Several studies have shown that hatchery-reared fish consume less food and fewer prey types, lag behind wild fish in switching to new prey, and experience depressed growth and survival soon after release compared with wild fish (reviewed by Olla et al. 1998). In a laboratory study hatchery-reared brown trout ate less, were slower to attack prey and were less efficient in consuming live novel prey compared with wild conspecifics (Sundström & Johnsson 2001), but foraging success improved with successive trials. In support, Johnsen & Ugedal (1986, 1989, 1990) showed that hatchery-reared brown trout initially ate fewer and different prey items compared to wild fish, and that these differences disappeared after a few weeks in the wild. In the present study, the initially lower growth rate of hatchery vs. wild trout may thus be explained by a lack of experience of feeding on live prey, whereas the more similar growth rates of wild and hatchery trout during the second growth period may reflect a gradual acclimation to natural conditions (Fig. 4). The initially poor performance of released hatchery fish in the present and other studies may also be due to the prior-residence effect. Previous work on brown trout fry suggests that a body weight advantage of 30% is needed to balance the advantage of prior residence in territorial conflicts between wild fish (Johnsson et al. 1999). In the present study, hatchery trout were on average 114% larger by weight than wild fish (Fig. 1), but the competitive advantage of larger size (Riechert 1998) may have been partly compensated by the prior residence advantage of the resident trout.

Hatchery trout moved more than wild fish (Fig. 4). Weiss & Schmutz (1999) reported that about half of the stocked domesticated fish (2–3 years old) they studied had moved outside the 200 m section where they were stocked. Young-of-the-year hatchery and domesticated trout stocked in a Danish stream tended to move more, up to 600 m, in a predominantly downstream direction, compared to introduced wild trout (Jørgensen & Berg 1991). Nevertheless, most trout were found at the same site almost a year later, indicating a skewed distribution of movements.

To summarize, our results show that population density affects growth in trout parr (yearlings and older). Thus, competition and possibly population regulation is apparently not limited to early life-stages (underyearlings) as has previously been suggested. Furthermore, density-dependent effects on resident fish were independent of whether competitors were wild or hatchery trout. Our findings suggest that the stocking of hatchery fish may negatively affect natural populations through competition.

Acknowledgements

We thank Uno Unger, Riitta Kallio, Kristina Johansson, Eino Niiranen, Ylva Fägerås, Harriet Eeley and Michael Lawes for valuable help in the field and the Båkab Fish Farm for caring for the fish in the hatchery. This work has been conducted with financial support from the Commission of the European Communities, Agriculture and Fisheries (FAIR) specific RTD programme, CT-97–3498, ‘Performance and Ecological Impacts of Introduced and Escaped Fish: Physiological and Behavioural Mechanisms’. This study does not necessarily reflect the views of CEC and in no way anticipates the Commission's future policy in this area. The experimental work was approved by the Ethical Committee for Animal Research in Göteborg (licence 136·96). Constructive criticism from two anonymous referees is gratefully acknowledged.

Ancillary