1Several studies have offered evidence for the occurrence of density-dependent growth in stream-living Brown Trout. However, such evidence has been gleaned for low-density populations, whereas studies on persistently high-density populations have claimed that growth is density-independent. Such a paradoxical observation is shared with other salmonids and has been assumed by several authors to suggest that stream salmonid populations may be regulated by two different mechanisms: density-dependent growth at low densities and density-dependent mortality, in the absence of density-dependent growth, at high densities.
2This comparative long-term study explored the occurrence of density-dependent growth by examining growth during the lifetime across cohorts in three stream-living Brown Trout populations representing the opposite extremes of growth and density documented throughout the species’ distributional range.
3This comparison highlighted identical growth–recruitment patterns in a high-density population with low potential for growth, in a low-density population with high potential for growth and in a population with intermediate traits. In the three populations, growth declined with increased recruitment describing negative power trajectories. These observations are consistent with there being a single, negative power relationship between growth and density where the effects of density dependence are stronger at low densities and become negligibly low at high densities.
4Stream-living Brown Trout populations may be regulated by the continuous operation of density dependence on growth and mortality. In poorly recruited cohorts density dependence may operate on growth but not on mortality during a time period after which density dependence operates on both growth and mortality. In highly recruited cohorts, density dependence operates simultaneously on growth and mortality from the youngest life stages.
The negative feedback nature of density-dependent growth along with its overwhelming effects on major life-history traits such as age at maturity and fecundity (Rose et al. 2001) and its potential to be translated into density-dependent mortality is deemed to be a major mechanism underlying the numerical regulation of fish populations (Lorenzen & Enberg 2001). Succinctly, among the numerous factors that may affect growth throughout the lifetime (Lobón-Cerviá 2005a), the operation of density dependence predicts depressed growth at high densities caused by decreased food intake due to competition when resources become depleted by the increased abundance of individuals (Heath 1992).
Such a paradoxical observation is shared with other stream salmonids (Ricker & Foerster 1948; McFadden, Alexander & Shetter 1967) but clashes with predictions from the negative feedback theory that states that the operation of density dependence should preferentially occur in high-density populations and/or its effects should be stronger the higher the population density. Remarkably, the nonoccurrence of, the elusive behaviour, or failures in detecting density-dependent growth in high-density populations where mortality is typically density-dependent (Elliott 1994) has led several authors to conclude that density dependence operates on growth primarily at low densities (Imre et al. 2005) and have further suggested that stream salmonids may be regulated by two different mechanisms: density-dependent growth at low densities and density-dependent mortality, in the absence of density-dependent growth, at high densities (Elliott 1994; Imre et al. 2005).
Nevertheless, two recent studies cast doubt on the nonoccurrence of density-dependent growth in high-density populations. Jenkins et al. (1999) and Lobón-Cerviá (2005a) hypothesized that, in the wild, persistently high-density populations concur with low potential for growth and elusiveness in the operation of density dependence may be caused by minor variations in growth over high mean densities to the extent that growth may become insensitive to increased densities. Therefore, the effects of density dependence may go undetected when the resolution of the data set is not sufficient to assess the contribution of interannual variations in growth (Jenkins et al. 1999; Lobón-Cerviá 2005a).
This study tracks that hypothesis by exploring the occurrence of density-dependent growth in three contrasting populations of stream-living Brown Trout located at the opposite extremes of growth and density reported throughout the species’ distributional range. The analysis of long-term variations in growth and density in a persistently high-density population with minor potential for growth (Bisballe Baeck, Denmark), in a low-density population with strong potential for growth (Rio Chaballos, Spain) and in a population with intermediate traits (Rio Choudral, Spain) permitted the testing of the following predictions: (1) density dependence does operate on growth in high-density populations; (2) at a single population scale either in low, middle or high ranges, the effects of density dependence are stronger at low densities resulting in negative power trajectories; however, (3) the effects of density dependence measured as the rates at which growth declined with increased density are stronger in persistently high-density populations.
Study area, materials and methods
the brown trout populations
The Brown Trout populations of this study and their home streams differ in essential features. Bisballe Baeck is a small, 450-m long stream flowing into Dollerup Baek before it enters into Lake Hald (central Jutland, Denmark). Bisballe Baeck is a nursery for the Lake Hald-dwelling population from where large size (30–50 cm) spawners immigrate into the stream every winter. A 170-m long stream section (area = 146 m2; width = 0·15–1·5 m; depth = 5–20 cm) located at mid-distance from the stream mouth was selected to track long-term trends in growth and density (Mortensen 1985).
Rio Chaballos and Rio Choudral are located in Asturias (north-western Spain). These two stony bottom streams of rapid water flow about 20 km apart in the opposite direction across the rugged and mountainous landscape of the Rio Esva drainage. In these two streams, Brown Trout are resident and complete their life cycle within small stream reaches (Lobón-Cerviá 2000, 2004). The study site in Rio Chaballos is one of the four sites (site S1) previously reported by Lobón-Cerviá (2005a,b). This is a 75-m long stream section (area = 250 m2; width = 3–5 m; depth = 20–50 cm) located at 4 km from the stream mouth. Rio Choudral is about half the size and discharge. In this stream, the study site is 100-m long with an area = 120 m2, width = 0·8–1·3 m, and depth = 10–30 cm.
In the three streams temporal variations in Brown Trout size and density were tracked by applying the three-pass removal method with electrofishing techniques (Lobón-Cerviá 1991). Monitoring of Bisballe Baek Brown Trout commenced at the end of March to early April and continued every 30 days (range 25–45 days) during the years 1974–92. Monitoring of the resident populations commenced in mid-May and has been carried out every second month since 1986 in Rio Chaballos and every fourth month since 1990 in Rio Choudral. During the early stages of these studies, scales were used for age determinations. In later years, individuals were assigned to age classes easily detectable in the length–frequency distributions, and scales were only used to confirm the age of a few doubtful individuals. Densities were estimated separately for age-0, age-1 and older individuals and are expressed as individuals m−2. In Bisballe Baeck, Brown Trout residency spanned for an average of 1250 days after emergence (Lobón-Cerviá & Mortensen 2005). In the two Spanish populations, life spans were 900–1000 days.
Mortality patterns differed markedly between the northern, lake-migratory and the two southern, resident populations. In the former, mortality was severe soon after emergence and density-dependent during the whole lifetime (Lobón-Cerviá & Mortensen 2006). In contrast, in the two resident populations no mortality occurred during more than half the lifetime (Lobón-Cerviá 2003, 2005b). The onset of mortality in the second summer of life occurred when individuals attained 13–14 cm in size (Rincón & Lobón-Cerviá 2002) and was density-dependent during the rest of the lifetime (Lobón-Cerviá 2005b).
I explored 20 data sets encompassing juvenile mass and density during the first summer of life of stream-living Brown Trout. This analysis permitted me to set the magnitudes and ranges of growth and density in the populations of this study within the ranges reported for the species’ distributional range. This large-scale data set is an extension of that previously used by Lobón-Cerviá (2005a) and encompassed unpublished data (J. Lobón-Cerviá, unpublished) and those collected from the literature. For this comparison, I focused on populations inhabiting streams of comparable size (width = 1–5 m; mean depths < 30 cm) where juvenile size and density were collected over periods = 9 years. This data set included populations belonging to the major life-history modes exhibited by Brown Trout: resident, anadromous, resident and migratory, and lake-migratory. The only exceptions concerned the experimental data reported by Crisp (1993) and Jenkins et al. (1999). However, these two data sets were included because they represent two consistent instances of density-dependent growth reported in the literature.
This data set encompassed the following populations: Rio Chaballos (Spain, four sites, C1–C4; Lobón-Cerviá 2005a); Rio La Viella (Spain, three sites V1–V3; Lobón-Cerviá 2005a); Rio Castañedo (Spain, three sites, A1–A3, Lobón-Cerviá unpubl.); Rio Choudral (Spain, two sites, O1–O2; Lobón-Cerviá, unpubl.); Shelligan Burn (Scotland, several sites pooled, SB; Egglishaw and Shackley 1977); Convict Creek (California, three sites, J1–J3; Jenkins et al. 1999); Slapestone and Dubby Sikes (England, two sites, S1–S2, Crisp 1993); Black Brows Beck (England, one site, BBB, Elliott 1984a,b) and Bisballe Baeck (Denmark, one site BB, this study).
Studies of density-dependent growth have typically quantified variations in growth rates or body mass vs. intracohort or intercohort densities (Grant & Imre 2005). However, the strong dependence of growth rate on initial size (Elliott 1994) and differential responses of growth and mortality across temporal scales may combine to obscure actual growth–density relationships. Therefore, I focused on functional relationships between recruitment and the growth of survivors in each cohort. I asked whether growth–recruitment relationships are well described by straight lines, or might be better described by power functions.
For this purpose, recruitment (Rc, ind m−2) or the amount of juveniles commencing a new cohort was assumed to be the abundance of 1-month-old individuals (the first numerical census of each cohort). Two different measures were used to quantify the growth of survivors in each single cohort throughout each lifetime: (1) the mean mass of individuals (WM, g) that is, the mean mass averaged from the second sample (i.e. once recruitment was excluded) until the last recorded individuals just before the total disappearance of that cohort. (2) Per capita production (PP, g) that reflects the mean growth in mass attained by an average survivor during a lifetime. The production rate of a cohort is by definition (Ivlev 1966; Chapman 1978) the total amount of fish tissue formed during the lifetime including that formed by individuals who do not survive to the very end of the cohort. For each single cohort, production rates were calculated for each time interval between two successive quantifications of mass and numbers starting in the second quantification (i.e. once recruitment was excluded) up to the total disappearance of that cohort. For every time interval I calculated (1) the initial biomass (B1, g m−2) as the mean mass (g) × density (ind m−2) and the mean biomass as B1 + B2/2 where B1 and B2 are the initial and final biomass of that time interval; (2) daily instantaneous growth rates as log(W2/W1)/t where W1 and W2 are the mean mass at the beginning and end of that time interval and t, the number of days; (3) production rates for those time intervals as the mean biomass × growth rate × number of days; (4) total cohort production rate as the sum of those production rates over the lifetime; and (5) per capita production as the total cohort production rate/mean density or the density of that cohort averaged from the second census to the total disappearance.
Only complete cohorts from emergence to total disappearance were used to test recruitment dependence hypotheses. Thus, this analysis encompassed 15 cohorts hatched from 1974 to 1988 in Bisballe Baeck and 14 cohorts (1986–99) and 10 cohorts (1990–99) in Rio Chaballos in Rio Choudral, respectively. The effects of recruitment as a predictor variable (X) on the two growth measures as response variables (Y) were explored by fitting straight lines and negative power functions. The Akaike information criterion (Burnham & Anderson 1998) was used to select the best model between the two models to fit the data. The Akaike weights (W) were calculated for the linear (Wlinear) and negative power (Wpower) models. The latter was assumed to be the best model to fit the data when the difference Wpower − Wlinear > 2, which is equivalent to a 73% probability that the model selected is correct (Motulsky & Christopoulos 2004).
setting spatiotemporal variations in growth and density across populations
An early exploration of the 20 data sets revealed a bewildering range of variation in the juveniles’ summer density (0·01–9·05 ind m−2) and body mass (0·7–14·3 g). A plot of juvenile mass at the end of the first summer growth (August/September) vs. the mean summer density (April/May–August/September) recorded for the 20 data sets encompassing 217 data pairs revealed that fish mass described a negative power trajectory over the whole range of densities with no population showing densities > 2 m−2 and fish masses > 5 g (Fig. 1a). Nevertheless, populations differed significantly in both fish mass and density (anova, F38,392 = 23·3, P < 0·001) and each single population occupied a specific range within that trajectory suggesting that the ranges of temporal variation were population-specific. Typically, resident Brown Trout with lower densities and greater potential for growth were located at the steep, left-side limb whereas sea-trout and lake-migratory trout showed markedly greater densities and lower potential for growth and were located at the opposite flat, right-side wing of the trajectory (Fig. 1a).
This trajectory is strongly consistent with Jenkins et al.'s (1999) and Lobón-Cerviá's (2005a) hypotheses where elusiveness of density dependence on growth in persistently high-density populations may be caused by minor variations in fish mass with increased density to definitively obscure the effects of density dependence. At the opposite extreme of that trajectory, a prompt detection of density-dependent effects would be expected in low density, high growth populations where fish mass declines sharply with minor increments in density.
The three populations selected for this study represent the opposite ends of that trajectory (Fig. 1b). At one extreme, fish mass in Rio Chaballos showed a steep decline in mass within a broad range (7·3–14·3 g) with minor variations in density (0·02–0·9 m−2). At the opposite extreme, fish mass in Bisballe Baeck varies within a very narrow range (1·4–3·1 g) over a remarkable range of densities (0·8–9·0 m−2) suggesting that the data set of this population is insufficient to detect the operation of density dependence. Similarly, fish mass and density in Rio Choudral located at an intermediate position exhibited ranges of variation probably too low to permit the detection of growth–density relationships (Fig. 1b).
Nevertheless, the two measures of growth indicated that in the three populations both mean mass (WM) and per capita production (PP) declined with increased recruitment (Rc) over the whole range of recruitment magnitudes. In six instances, the Akaike weights computed for the negative power trajectories were > 3 relative to those obtained for linear regressions, leading to the selection of a negative power function as the best model to fit the six trajectories.
Negative power functions fitted to mean mass and per capita production over recruitment were all highly significant and strongly consistent with each other. Year-to-year variation in recruitment explained 48·2% (Rio Chaballos), 74·3% (Rio Choudral) and 74·1% (Bisballe Baeck) of the variations in per capita production and 57·3% (Rio Chaballos), 46·2% (Rio Choudral) and 42·9% (Bisballe Baeck) of the variations in mean mass (Fig. 2). Moreover, ancovas for differences among slopes for the two sets of three regressions (per capita production vs. recruitment and mean mass vs. recruitment) revealed highly significant differences for the log10-transformed data. For the set of three PP-Rc relationships, overall R2 = 0·96, P < 0·001 and F2,33 = 5·45, P = 0·01; and for the set of three WM-Rc relationships, overall R2 = 0·89, P < 0·001 and F2,33 = 4·72, P = 0·02. The corresponding Tukey test indicated steeper slopes in Bisballe Baeck > Rio Choudral > Rio Chaballos.
In the three populations growth declined with increased recruitment and described negative power trajectories. However, the rates at which growth declined per unit of density increases were greater in the persistently high-density population with low potential for growth, markedly smaller in the lowest density population with stronger growth potential, and intermediate in Rio Choudral exhibiting intermediate traits.
In response to the elusive behaviour of density dependence on growth in high-density populations, Jenkins et al. (1999) and Lobón-Cerviá (2005a) hypothesized that its operation may be obscured when observational data are not sufficient to capture the contribution of interannual variation in growth. The comparison of fish mass at the end of the first summer growth vs. mean density across 20 data sets supported this claim. Therefore, this study focused on the growth experienced by survivors during the whole lifetime vs. the initial abundance of recruits commencing that cohort (i.e. recruitment). This approach has shown to be most appropriate to elucidate the occurrence of density-dependent growth in populations differing markedly in growth and density.
This among-population comparison has provided evidence for the occurrence of density-dependent growth in the greatest density population with minor potential for growth recorded throughout the species distributional range. Remarkably, the growth–density pattern elucidated for this population was identical to those elucidated for low and mid-density populations with stronger potential for growth. In the three populations, faster growth in individuals of poorly recruited cohorts slowed down in highly recruited cohorts and the temporal variations in growth described negative power trajectories over density. Strong consistency among patterns emphasize the importance of the initial number of individuals commencing a cohort as a major factor influencing growth whose effects remain beyond the effects exerted by the numerous other factors actually operating on growth throughout the lifetime (Lobón-Cerviá 2005a). Moreover, the negative power trajectories detected for the three populations are consistent with the prediction that on a single population scale, the effects of density dependence are stronger at low densities. and such effects, as measured by the slope of the growth–density relationships, are stronger in persistently high-density population. This is, in turn, consistent with the negative feedback theory predicting an increased chance for the occurrence of, or for stronger effects exerted by, density dependence in high-density populations.
These findings combine with other observational (Jenkins et al. 1999; Lobón-Cerviá 2005a) and experimental studies (Bohlin et al. 2002) to support the occurrence of density-dependent growth as a common process and underscore the importance of density-dependent growth as a major factor determining size-at-age in stream-living Brown Trout. Consequently, the paradoxical observation that density dependence operates on growth in low-density but not in high-density populations, seems more apparent than real. This may have been caused by methodological drawbacks such as inadequate data sets of low resolution, growth measures unsuitable to detect variation, or serial data sets where the time periods analysed had not been long enough to capture sufficient temporal variability.
Growth and mortality are deemed major regulatory processes in stream Brown Trout (McFadden & Cooper 1964; Elliott 1994) and other salmonids (McFadden et al. 1967; Imre et al. 2005; Quinn 2005), upon which density dependence may operate and translate its effects into each other process. Failures to detect density-dependent growth in high-density populations where mortality is typically density-dependent has led several authors to suggest that these populations may be regulated by density-dependent growth at low densities (Imre et al. 2005) and by density-dependent mortality, in the absence of density-dependent growth, at high densities (Elliott 1994; Imre et al. 2005). Nevertheless, Lobón-Cerviá (2005a) hypothesized that if the mechanism underlying density-dependent growth is decreased food intake due to competition when resources become depleted (Heath 1992) then density dependence should operate on growth prior to operating on mortality. Rincón & Lobón-Cerviá (2002) and Lobón-Cerviá & Mortensen (2006) examined the populations of Rio Chaballos and Bisballe Baeck and suggested that spatial habitat availability is size-selective and becomes progressively limiting for growing individuals. Under increasingly limited space, smaller sizes induced by the operation of density dependence may be advantageous for survival. Thus, the role of the prior operation of density dependence on growth would be to prevent mortality so as to maximize survival for maintaining the population at the highest possible abundance.
The patterns of density-dependent growth and mortality elucidated for Rio Chaballos (Lobón-Cerviá 2005a,b), Rio Choudral (Lobón-Cerviá unpubl.) and Bisballe Baeck (Lobón-Cerviá & Mortensen 2005, 2006) support this hypothesis. For the very same study site of Rio Chaballos, Lobón-Cerviá (2005a, b) documented density-dependent growth during a prolonged time period of negligible mortality that lasted for more than half a lifetime. As mortality occurred by the end of the second summer, density dependence operated simultaneously on both growth and mortality. Unlike in this stream, severe mortality occurred soon after emergence in the high density, slow growth population of Bisballe Baeck (Lobón-Cerviá & Mortensen 2006). Such severe mortality, in combination with a weak potential for growth, might obscure the operation of density dependence on growth but, as inferred from this study, the operation of density dependence on mortality throughout the lifetime concurs with the operation of density dependence on growth.
Overall, these patterns suggest that stream-living Brown Trout populations may be regulated by the continuous operation of density dependence on growth and mortality where density-dependent adjustments of population numbers through density-dependent mortality should invariably be associated with depressed density-dependent growth (Lobón-Cerviá 2005b). In poorly recruited cohorts mortality may be weak or negligible during an earlier time period in which density dependence only operates on growth. During a second time period when mortality occurs, density dependence operates on both growth and mortality. In highly recruited cohorts, density dependence operates simultaneously on growth and mortality since the youngest life stages.
The Bisballe Baeck data set was collected by the late Dr Erik Mortensen when he served as a senior researcher for the National Environmental Research Council of Denmark. Data collections for Rio Chaballos and Rio Choudral were financed by the Spanish MEC-CICYT projects no. PB84-0142, PB89-0048, and PB92-0045 and by the Municipality of Valdés (Asturias). Permission for fishing was by agreement between CSIC and Principado de Asturias. Comments by Dr A. Vollestad, Dr T. Northcote, and three anonymous reviewers improved an earlier draft. Ms Maite Lavandeira improved the English style.