Ice-cover effects on competitive interactions between two fish species


Correspondence author. E-mail:


1. Variations in the strength of ecological interactions between seasons have received little attention, despite an increased focus on climate alterations on ecosystems. Particularly, the winter situation is often neglected when studying competitive interactions. In northern temperate freshwaters, winter implies low temperatures and reduced food availability, but also strong reduction in ambient light because of ice and snow cover. Here, we study how brown trout [Salmo trutta (L.)] respond to variations in ice-cover duration and competition with Arctic charr [Salvelinus alpinus (L.)], by linking laboratory-derived physiological performance and field data on variation in abundance among and within natural brown trout populations.

2. Both Arctic charr and brown trout reduced resting metabolic rate under simulated ice-cover (darkness) in the laboratory, compared to no ice (6-h daylight). However, in contrast to brown trout, Arctic charr was able to obtain positive growth rate in darkness and had higher food intake in tank experiments than brown trout. Arctic charr also performed better (lower energy loss) under simulated ice-cover in a semi-natural environment with natural food supply.

3. When comparing brown trout biomass across 190 Norwegian lakes along a climate gradient, longer ice-covered duration decreased the biomass only in lakes where brown trout lived together with Arctic charr. We were not able to detect any effect of ice-cover on brown trout biomass in lakes where brown trout was the only fish species.

4. Similarly, a 25-year time series from a lake with both brown trout and Arctic charr showed that brown trout population growth rate depended on the interaction between ice breakup date and Arctic charr abundance. High charr abundance was correlated with low trout population growth rate only in combination with long winters.

5. In conclusion, the two species differed in performance under ice, and the observed outcome of competition in natural populations was strongly dependent on duration of the ice-covered period. Our study shows that changes in ice phenology may alter species interactions in Northern aquatic systems. Increased knowledge of how adaptations to winter conditions differ among coexisting species is therefore vital for our understanding of ecological impacts of climate change.


Interactions between species shape climate impacts on organisms at various scales and are important to consider when assessing ecological responses to climate change (Gilman et al. 2010). However, although the timing and phenology of life-history events have received increasing attention, less has been given to the seasonal succession in interspecific interactions (Yang & Rudolf 2010). Particularly, for animals living in seasonal temperate environments, observed and expected changes in snow and ice conditions are one of the most pronounced impacts of global warming (Smol et al. 2005; Kausrud et al. 2008). The winter is usually a critical period because of harsh climatic conditions and reduced resource availability (Hurst 2007; McNamara & Houston 2008). Nonmigratory animals that do not hibernate, but remain active during winter face physiological constraints that often lead to a negative energy budget. Thus, to survive, they frequently depend on stored energy reserves in addition to feeding (Bull, Metcalfe & Mangel 1996; De Block, McPeek & Stoks 2007). The overwinter survival in a population may vary between years as a response to inter-annual variations in climatic conditions, such as the amount of snow and ice, or length of the winter season (i.e. duration of periods with frost, ice or snow cover). Ice formation and deep snow decreases the access to food (Solberg et al. 2001; Doherty & Grubb 2002; Robinson, Baillie & Crick 2007), but may also function as a shield from predators (Finstad et al. 2004; Berg et al. 2008). Variations in snow cover explained up to 65% of inter-annual dynamics of four resident terrestrial species at Greenland (Forchhammer et al. 2008). Although a number of studies have focused on how predator–prey relationships fluctuate with variations in seasonal timing and phenology (Thompson & Townsend 1999; Parmesan 2006), less attention has been given to the effect of winter length on competitive interactions.

Seasonal reduction in resource availability can cause increased competition among coexisting species (Schoener 1982). Additionally, populations may show adaptations that influence their competitive ability during different seasons. Some species are specialised to succeed under winter conditions, for example the carnivore wolverine (Gulo gulo) is adapted to feeding when there is snow cover, and may therefore benefits from winters with large amounts of snowpack (Brodie & Post 2010). Variation in cold tolerance and adaptations to winter determines species distribution among passerine birds (Swanson & Garland 2009). Atlantic salmon (Salmo salar) populations show adaptive variation in feeding activity along a climatic gradient, whereby northern populations originating from areas with longer winters have higher feeding activity with decreasing lipid-depletion rate during winter, a response not present in southern populations (Finstad et al. 2010a).

In aquatic ecosystems, variations in duration of the ice-cover may have large impacts on species and their interactions by changing important factors like light and temperature (Adrian et al. 1999; Christoffersen et al. 2008). Here, we study the interactions between two competing salmonid fish species during the winter season. Brown trout [Salmo trutta (L.)] and Arctic charr [Salvelinus alpinus (L.)] coexist in relatively simple communities along a wide climate gradient, and variation in competitive interactions result in a diversity of dominance ratios and population size structures (Svärdson 1976; Finstad et al. 2010b). The two species often segregate in diet and habitat, probably resulting from both exploitative and interference competition. Typically, brown trout is more aggressive and forces Arctic charr away from the most profitable habitat in the littoral zone into the pelagic or deeper water (Langeland et al. 1991; Jansen et al. 2002; Forseth et al. 2003). Thus, brown trout is considered a stronger competitor than Arctic charr during summer, but less is known on how their ecological interactions change with season. Arctic charr is the world’s most northerly distributed freshwater fish and hence expected to be adapted to winter conditions (Klemetsen et al. 2003a; Siikavuopio et al. 2009). We therefore predicted that long winters (i.e. long duration of the ice-covered period) will favour Arctic charr over brown trout and influence their ecological interactions in lakes where they coexist. This prediction is tested by linking laboratory-derived physiological performance to abundance parameters from field studies. Metabolism, growth and food consumption under simulated ice-cover are studied for both species in experimental settings. Furthermore, to study the effect of competition with Arctic charr on brown trout in natural systems, brown trout abundance is compared both spatially (190 lakes with and without Arctic charr) and temporally (25 years in a lake with coexisting brown trout and Arctic charr), by assessing the combined effect of presence of Arctic charr and length of the ice-covered period.

Materials and methods

Physiological performance under simulated ice-cover

Ice formation causes accumulation of snow on the top of the lake, and a layer of snow will cause up to 99% of the light to be attenuated (Wetzel 2001; Jewson et al. 2009). Therefore, light effects on metabolic rate, growth and food consumption were measured by simulating ice-cover or no ice-cover in two similar tank compartments with different light treatments, but similar temperature. In addition to these laboratory experiments, outdoor semi-natural stream channels were used to study the effect of ice-cover on the energy use and feeding under conditions more similar to natural environments. Arctic charr and brown trout used were one-summer-old (0+), first-generation hatchery-reared individuals with parents originating from the river Imsa catchment in south-western Norway (58°90′N, 5°90′E). Mean weight of fish was 18·2 g (±2·32 SD) for brown trout and 29·1 g (±6·14 SD) for Arctic charr. Mass differences were accounted for by mass standardising growth and food consumption rates according to published allometries (e.g. Elliott & Hurley 1997, 1998) or using mass as covariate in the statistical analyses. All experiments were conducted at the Norwegian Institute of Nature (NINA) Research Station Ims from January to March 2004. Further details on the methodology of the metabolic rate measurements, tank and outdoor channel experiments can be found in Finstad et al. (2004, 2007) and Finstad & Forseth (2006), describing the same set-up used on Atlantic salmon.

In the laboratory experiments, two compartments light isolated as in a photolaboratory darkroom were used, either given an ice-cover treatment (darkness) or no ice-cover [6-h light (≈70 lux)/18-h dark] as explained by Finstad et al. (2004). In ice-covered lakes, the temperature varies between 0 and 4 °C, but 4 °C is only found in the deeper areas (Wetzel 2001). Hence, temperature was maintained at ≈1 °C during all laboratory experiments and was similar among light treatments (mean 1·3 °C ± 0·5 SD).

Resting metabolic rates (Jobling 1994) were measured in both light treatments using a flow-through respirometer as described by Finstad et al. (2007), based on a commonly used protocol (McCarthy 2001; Álvarez, Cano & Nicieza 2006; Millidine, Armstrong & Metcalfe 2006). Metabolic rate was determined by placing individual fish inside a respirometer tube (180 × 50 mm) and measure the oxygen concentration in the inflowing and outflowing water using a microcathode oxygen electrode (model 1302) connected to an oxygen meter (model 781) (Strathkelvin Instruments Ltd., Glasgow, Scotland). The flow through the chamber was measured by weighing the amount of water (±0·001 g) from the outlet of the tubes for a period of minimum 60 s. A maximum of 20% reduction in oxygen concentration was allowed, and the flow was held constant (mean 60 mL min−1 ± 18 SD) throughout both measurements and acclimation periods. The fish were kept in experimental tanks and acclimated to light or dark treatment for minimum three weeks before metabolic measurements. They were deprived of food for the last ≈48 h before being introduced to the tubes. After one night acclimation in the tubes (minimum 12 h), three measurements of oxygen consumption were obtained for each fish, and average metabolic rate calculated. The oxycalorific coefficient was set to 13·59 kJ g−1 (Jobling 1994). For each species, 15 individuals and one empty chamber (control) were measured in both treatments. To ensure that only inactive fish were included, the ten fish with the lowest oxygen consumption was used. The fish used for metabolic measurements were not reused for further experiments. The effect of light treatment and species on ln-transformed metabolic rate was tested using ancova with ln-transformed mass as covariate.

Growth rates and food consumption in tank environment were compared by running two replicates in each light regime for each of the two species. Experimental units were randomly distributed to avoid systematic tank effects. The tanks were 45 × 45 and 60 cm deep and had a water flow 2 L per min and water depth of 30 cm. In each of the eight tanks, 10 individually marked fish were used. All tank experiments lasted for 48 days, and no fish died. Growth was measured as the standardised mass-specific growth rate (Ω%) (Ostrovsky 1995)

image(eqn 1)

where M0 and Mt are the respective body mass (g) at the beginning and end of each experiment, t is the experimental period (days) and b is the allometric mass exponent for the relation between specific growth rate and body mass fixed at 0·3 for both species (Elliott & Hurley 1997; Larsson et al. 2005). Food consumption was estimated using a caesium (Cs) tracer methodology (Forseth et al. 2001; Jonsson et al. 2001; Finstad et al. 2004). The fish were fed in excess with CsCl-enriched granulated fish food (Felleskjøpet, Sandnes, Norway, Cs concentration: 14·1 ppm fresh mass) administered from automatic feeders. Measured food consumption is based on estimating the intake of Cs from an observed change in Cs body burden with time. Based on known rates of assimilation and elimination, the food consumption is obtained by dividing the Cs intake by the concentration in food (Forseth et al. 1992, 2001). Species differences in light response of growth and food consumption was tested using two-way anova. In accordance with Underwood (1997), post hoc pooling of replicates (removal of the tank factor from the model) were conducted for α larger than 0·25.

To study performance with or without ice-cover under more natural conditions, outdoor semi-natural channels (485 × 50 cm, water depth ≈30 cm) were used in addition to the tank experiments (Finstad et al. 2004). Working with natural-ice-cover in a controlled setting was not possible because of the logistically difficulties with handling water at freezing temperatures. Ice-cover was therefore simulated by covering the outdoor channels with light-impermeable material. No ice-cover was simulated by covering the channels with clear plastic to prevent drift of exogenous material into the system while allowing natural daylight. All outdoor channels received the same inflow of water from a nearby natural lake, ensuring identical temperatures and food conditions for both ice-cover treatments. No additional food other than invertebrate prey naturally present in the intake water was given. Mean temperature was 2·5 °C (±0·3 SD), and water flow through each channel was ≈50 L min−1. For each treatment and species, three replicates with ten fish in each cannel were used. To avoid possible confounding effects of olfactory cues different species were not mixed within channels. We were therefore not able to test for ice × species interactions, but used a design with ice-cover effect nested within channel and species. The study lasted for 54 days and in total, 18 of 120 fish died during the channel experiment. Brown trout and Arctic charr usually deplete energy reserves during winter in their natural environment (Berg, Thronæs & Bremset 2000; Finstad, Berg & Lohrmann 2003). Individual energy loss rates (E) were calculated as

image(eqn 2)

where inline image(see eqn 1 for explanation of terms) and J is the mass-specific energy (J g−1). J was determined using wet weight/dry weight ratio as proxy for energy content as described by Finstad & Forseth (2006).

Brown trout biomass in presence and absence of arctic charr

Data on brown trout catches were extracted from a data base containing results from test fishing of more than 400 Norwegian brown trout populations carried out by Norwegian management and research institutions in the period 1972–1997 (Ugedal, Forseth & Hesthagen 2005). All lakes were fished with a standardised series of gillnets consisting of eight nets (25 × 1·5 m) with mesh sizes from 21 to 52 mm (knot to knot) (Jensen 1977). This gillnet series captures brown trout between 190 and 500 mm body length in a comparable way. The nets were distributed along the shoreline, and the lakes were fished during summer, with different effort (i.e. number of gillnet series) depending on lake size. Catch per unit effort (CPUE) based on total weight of the brown trout catch (weight per unit catch effort (WPUE) in gram) per 100 m2 gillnet area per night was used as a proxy for biomass. Biomass was considered a better measure of population size than number of individuals when comparing brown trout populations between lakes, because of differences in productivity and environmental conditions across lakes which may cause large variations in body size and number of individuals among populations.

To study the effect of competition with Arctic charr on brown trout in natural systems, lakes with only brown trout and those with brown trout living together with Arctic charr were compared. Information on presence or absence of charr was used because information on charr biomass was not available from the brown trout data base. Lakes with brown trout stocking and those containing other fish species were excluded. However, because of their common immigration history after the last glaciation, threespined stickleback (Gasterosteus aculeatus) co-occurs with brown trout and Arctic charr in most lakes below the marine transgression line (Huitfeldt-Kaas 1918). It was therefore necessary to include some lakes with threespined stickleback (20 lakes) to achieve sufficiently high number of data points at lower altitudes. This resulted in 144 lakes without Arctic charr and 46 lakes with coexisting brown trout and Arctic charr included in the analyses. The lakes are distributed all over Norway from 59° to 70°N (Fig. 1) and are located between 15 and 1466 m.a.s.l. They vary in size from 1 to 5500 ha. Lake morphology (area and perimeter) was extracted digitally for all lakes from maps in the program ArcGIS Desktop 9·3. Lake ice breakup was predicted based on the lake-specific annual mean air temperature normal from 1961 to 1990 (Tveito et al. 2000) as given by Weyhenmeyer, Meili & Livingstone (2004);

image(eqn 3)

where TB is the predicted median Julian day of ice breakup and Tm is the site-specific temperature mean. This prediction model was developed for Swedish (eastern neighbouring country of Norway) lakes between 55° and 70°N, and is thus expected to be comparable to Norwegian lakes at similar latitudes.

Figure 1.

 Geographical distribution of the 144 Norwegian lakes with only brown trout (grey) and the 46 lakes with brown trout and Arctic charr (black) included in the analyses. Lake Atnsjøen with 25 years of test fishing is denoted with an asterisk.

The combined effect of competition from Arctic charr and ice-cover duration on brown trout biomass was modelled with multiple regression treating presence or absence of Arctic charr as a binary explanatory variable and the predicted Julian day of ice breakup as a proxy for duration of the ice-covered period. To confirm that ice breakup date reflects variation in ice-cover duration the correlation between the two variables was tested (Pearson’s r = 0·612, d.f. = 211, P < 0·001) based on a independent data set containing observed ice duration data for eight U.S. lakes in the period 1982–2010 (available at, Ice duration Trout Lake Area, North Temperate Lakes Long Term Ecological Research program, NSF, E. Stanley, Center for Limnology, University of Wisconsin-Madison). Lake morphology likely influences both time of ice breakup and the probability for coexistence of brown trout and Arctic charr; hence, lake area was included in the global model. Larger lakes will usually be deeper and have a more complex bathymetry offering more niches for fish species. Lake perimeter was used as a proxy for the amount of the preferred littoral habitat for Arctic charr and brown trout (Langeland et al. 1991; Jansen et al. 2002). Additionally, two-way interactions between presence/absence of Arctic charr and each of these three explanatory variables were included. Of the total 190 lakes, 40 were affected by water level changes because of hydropower production. To account for any potential effect of hydropower production, a preliminary test was performed including hydropower as a binary explanatory variable (water level changes vs. no changes). Because this variable did not affect the main results, it was excluded from the final model. The trout biomass, lake area and lake perimeter were ln-transformed to stabilise the variance or reduce skew. To identify the most parsimonious model, all subset models were compared with the global model based on Akaike Information Criterion (AIC, Burnham & Anderson 2002). All statistics and modelling was performed in R software version 2.9.2 (The R foundation for Statistical Computing 2009).

Temporal variations in Brown trout population growth rate in presence of arctic charr

We further tested if the competitive influence of Arctic charr on brown trout is influenced by ice-cover duration, using a 25-year (1985–2009) time series of trout and charr abundance from standardised test fishing in Lake Atnsjøen, Norway (61°51′N, 10°13′E, 701 m.a.s.l., Fig. 1). Both fish species were sampled in August every year, using gillnets covering depths down to 75 m in the epibenthic zone and 12 m in the pelagic zone, as described by Hesthagen et al. (2004). Because the type of gillnet series used was changed in 1994 (see Hesthagen et al. 2004 for details), only fish collected in net panels with comparable mesh sizes (16–45 mm, knot to knot) during the full 25-year period were included in the analyses. CPUE of both brown trout and Arctic charr were calculated as total numbers of fish caught per 100 m2 gillnet area per 12 h fishing. Observed ice breakup dates for each year was obtained from the Norwegian Water Resources and Energy Directorate.

Population growth rate of brown trout (λ) was estimated using CPUE from gillnets as a proxy for population size;

image(eqn 4)

where CPUEt and CPUEt−1 are the gillnet CPUE in year t and year t−1, respectively. Brown trout CPUEt−1 was entered as explanatory variable because density-dependent effects of brown trout was assumed a priori. Spurious density dependence in the relationship between brown trout biomass and growth rate (sensuFreckleton et al. 2006) was not tested for; therefore, effects of brown trout CPUEt−1 should be treated with caution. However, because there was no correlation between Arctic charr and brown trout catches the same year (Pearson r = 0·019, P = 0·931), spurious density dependence would not be a problem when testing for Arctic charr density on brown trout growth rate. Autocorrelations were tested for by including autoregressive error structure, but without any evident improvement in model fit (AR-1 structure fitted with the nlme library of R (Pinheiro et al. 2009), AIC = 56·88; linear model without AR-1 structure, AIC = 57·98). Further modelling was accordingly carried out with ordinary multiple regression.

How the difference in ice-cover duration influenced the effects of Arctic charr population size on brown trout population growth rate (λ) was tested for by entering the interaction effect of Arctic charr population size in year t−1 (inline image) and the time of ice breakup (Julian date) in a linear model. Because of limited degrees of freedom, a full model selection procedure including all possible interactions was not possible. Therefore, the hypothesis that competitive influence of Arctic charr on brown trout is influenced by ice-cover duration was tested by looking at improvement in the model fit when including the interaction term between Arctic charr and ice-cover duration.


Physiological performance under simulated ice-cover

Resting metabolic rate of fish reared in darkness was lower than for fish reared in 6-h daylight for both species (Fig. 2a, ancova with ln mass as covariate, Arctic charr: ln mass, F1,17 = 0·17, P = 0·681; light, F1,17 = 9·69, P = 0·006, brown trout: ln mass, F1,17 = 3·05, P = 0·098; light, F1,17 = 6·01, P = 0·025). The effect of light treatment on metabolic rate did not differ between Arctic charr and brown trout (ancova: ln mass, F1,35 = 0·67, P = 0·415; species, F1,35 = 0·76, P = 0·389; light, F1,35 = 16·65, P < 0·001; light × species, F1,35 < 0·01, P = 0·925).

Figure 2.

 Energetic responses of Arctic charr and brown trout to experimentally induced ice-cover in laboratory (a–c) and in semi-natural outdoor channels (d). Mean (±SE) resting metabolic rate (a), mass standardised growth (b) and food consumption (c) for fish reared in darkness (black bars) or 6-h daylight (white bars) in laboratory. Lower panel (d) shows mean (±SE) mass standardised loss of energy in semi-natural outdoor channels with clear plastic (white bars) and with light-impermeable cover (black bars) during winter. Different letters (x, y, z) indicate statistically significant differences (P < 0·05, Tukey’s HSD multiple comparisons) between groups.

The effect of light treatment on mass standardised growth rates differed between species in the tank experiments (anova: light, F1,76 = 33·01, P < 0·001; species, F1,76 = 174·67, P < 0·001; light × species, F1,76 = 8·56, P = 0·004). While both species retained positive or neutral growth in the 6-h light treatment, only Arctic charr was able to sustain positive growth in the dark (Fig. 2b). Similarly, the effect of light treatment on mass standardised food consumption also differed between brown trout and Arctic charr (anova: light, F1,76 = 25·81, P < 0·001; species, F1,76 = 36·92, P < 0·001; light × species, F1,76 = 6·19, P = 0·015). Arctic charr had higher intake rates than brown trout in both light and darkness (Fig. 2c).

All fish reared in semi-natural outdoor channels had negative growth rates (Fig. 2d, mean energy loss Arctic charr 10·90 ± 3·8 SD J g−1day−1; brown trout 11·39 ± 3·7 SD J g−1 day−1). There was no effect of ice-cover treatment on brown trout (Welch t = −0·04, d.f. = 49·48, P = 0·963), while Arctic charr had 17% lower energy loss under simulated ice than without ice (9·88 ± 3·8SD and 11·93 ± 3·6 SD Jg−1day−1, respectively, t = 2·13, d.f. = 57·80, P = 0·036).

Brown trout biomass in presence and absence of arctic charr

Brown trout biomass decreased significantly with increasing ice-covered period in lakes with Arctic charr present (Fig. 3). In contrast, no effect of ice-cover duration on brown trout biomass was detected in lakes without Arctic charr. Variation in brown trout biomass was best explained by a model including presence or absence of Arctic charr, ice breakup date and the interaction between Arctic charr and ice breakup date (Table 1, for model selection details see Table S1 Supporting Information). Brown trout biomass was lower in presence of Arctic charr than in lakes where Arctic charr was absent (Fig. 3). Furthermore, there was no significant effect of ice breakup date on brown trout living without Arctic charr, but a significant negative effect of length of the ice-covered period when Arctic charr was present (Fig. 3, anova: charr, F1,186 = 39·54, P < 0·001; ice breakup, F1,186 < 0·01, P = 0·928; charr × ice breakup, F1,186 = 7·57, P = 0·006).

Figure 3.

 The relationship between brown trout biomass (WPUE in gram, ln-transformed) and ice-cover duration (Julian date of estimated ice breakup) for lakes without Arctic charr (grey circles, stippled line) and lakes with Arctic charr present (black circles, solid line). Lines indicate least square regression fit (Arctic charr present: y = 10·86 − 0·02x, r2 = 0·09, P = 0·042; Arctic charr absent: y = 6·93 + 0·01x, r2 = 0·14, P = 0·155). One lake with Arctic charr and one without were excluded from the analyses because they were considered outliers, but exclusion of these data points did not significantly affect the results.

Table 1.   Parameter estimates for best least square regression model between brown trout biomass in 190 Norwegian lakes, with presence of Arctic charr and ice breakup date as explanatory variables. The model was identified by comparing AIC-values of all possible subset models (see Table S1 Supporting Information for model selection)
CoefficientsEstimate (±SE)t-valueP
  1. Model statistics: F3,186 = 15·71, P < 0·001, r2 = 0·20.

Intercept10·8644 (±1·5520)7·000<0·001
Charr−3·9279 (±1·7833)−2·2030·028
Ice breakup−0·0286 (±0·0119)−2·3980·017
Charr × Ice breakup0·0371 (±0·0134)2·7530·006

Temporal variations in brown trout population growth rate in presence of arctic charr

The brown trout population in Lake Atnsjøen was negatively affected by the combination of high Arctic charr abundance and long ice-cover duration (Fig. 4). This was evident from the better model fit of the full model including the interaction between ice breakup and Arctic charr abundance (ΔAIC = 3·50). There was also an apparent negative correlation between brown trout population size the previous year and the growth rate (Table 2). A negative growth rate of brown trout population was found only when both Arctic charr density was high and the ice-cover lasted long (Fig. 4b). This was confirmed by a significant negative interaction between ice breakup date and Arctic charr population size the previous year (Table 2, anova: CPUE troutt−1, F1,18 = 22·86, P < 0·001; CPUE charrt−1, F1,18 = 0·03, P = 0·859; ice breakup, F1,18 < 0·01, P = 0·998; CPUE charrt−1 × ice breakup, F1,18 = 4·83, P = 0·042).

Figure 4.

 Effects of variation in Arctic charr abundance and ice breakup date on the brown trout population in Lake Atnsjøen in the period 1985–2009. (a) Catch per unit effort (CPUE) of Arctic charr (black) and brown trout (grey) each year. Open symbols and stippled line indicates Julian day of ice breakup (right y-axis). (b) Contour plot of the predicted brown trout population growth rate (λ) against CPUE of Arctic charr the previous year (CPUE charr lag 1 year) and Julian date of ice breakup. Darker colour indicates more negative growth rate.

Table 2.   Parameter estimates for best least square regression model between brown trout population growth rate (λ) in Lake Atnsjøen and the gill net catch per unit effort (CPUE) of brown trout and Arctic charr and Julian date for ice breakup as explanatory variables
CoefficientsEstimate (±SE)t-valueP
  1. Model statistics: F3,18 = 6·29, P = 0·004, r2 = 0·53.

Intercept−2·4820 (±1·8250)−1·3600·191
CPUEtroutt−1−0·0118 (±0·0022)−5·173<0·001
CPUEcharrt−10·0261 (±0·0120)2·1800·043
Ice0·0253 (±0·0132)1·9160·072
CPUEcharrt−1 × Ice−0·0001 (±0·0001)−2·1990·042


This study shows that contrasting performance during winter may shape the climate impact on competitive interactions between two salmonid fishes. Under winter conditions, Arctic charr had higher growth and food consumption rate than brown trout in laboratory environment, and the species differences were particularly pronounced in light conditions simulating those found under ice-cover. In more adverse semi-natural experimental arenas, Arctic charr responded positively (i.e. had smaller weight loss) to simulated ice-cover, whereas brown trout remained indifferent to ice-cover treatment. Using both spatial and temporal analyses from Norwegian brown trout populations, we show that the impact of Arctic charr on brown trout populations differ with varying ice-cover duration, probably due to these differences in physiological response to winter. Brown trout population size, measured as weight per unit catch effort as a proxy for biomass, showed a significant negative effect of late ice breakup when Arctic charr was present. In contrast, there was no significant effect of ice-cover duration on brown trout populations in lakes without Arctic charr. Similarly, in time-series analyses, there was a negative effect on brown trout population growth rate during periods with a strong Arctic charr population combined with late ice breakup. Put together, these results clearly indicate that physiological capabilities of Arctic charr to perform in low-light and cold environment make the species a stronger competitor when the lake is ice-covered and the environmental conditions are comparably more unfavourable for brown trout.

While Arctic charr is able to feed continuously in darkness during winter (Svenning, Klemetsen & Olsen 2007), brown trout has generally been regarded to be more dependent on light for feeding (e.g. Langeland et al. 1991; Klemetsen et al. 2003a). However, it has recently been showed that brown trout responds to very low light levels under experimental settings (Rader et al. 2007). Furthermore, in a field study in a subarctic lake, both brown trout and Arctic charr were found to feed actively during the ice-covered period, in spite of low temperatures and low light levels (Amundsen & Knudsen 2009). Even when feeding has been observed during winter, the consumption rates for both species have been considered too low for growth (Klemetsen et al. 2003b; Amundsen & Knudsen 2009). This is in accordance with our results for brown trout which indicate that winter acclimatisation reduces the feeding activity resulting in low food intake irrespective of light conditions and a negative growth rate under simulated ice-cover. Reduced feeding motivation during winter is also common among Atlantic salmon (Metcalfe, Huntingford & Thorpe 1986). However, our tank experiments confirmed that when food supply was sufficient, Arctic charr was able to sustain positive growth also in darkness, because of the relatively high food consumption under winter conditions. Under more natural low prey abundance in the outdoor channels, none of the two species were able to feed enough to avoid energy loss, but Arctic charr was able to minimise the loss under ice-cover. Hence, the combination of low metabolism, relatively high intake rate and high food conversion efficiency makes Arctic charr strongly adapted to winter conditions, and this species performs better under ice-cover than with no ice.

The ecological interactions between brown trout and Arctic charr vary between winter and summer (Amundsen & Knudsen 2009). The two species show similar habitat and dietary preferences in allopatry, but because brown trout is more aggressive, it usually dominates over Arctic charr in the preferred littoral habitat in lakes where the two species coexist. Hence, Arctic charr is mainly found in deeper water along the bottom or in the pelagic area during summer when living together with brown trout (Langeland et al. 1991; Jansen et al. 2002). Just before ice formation, however, Arctic charr performs a seasonal habitat shift and moves to the same habitat as brown trout (Klemetsen et al. 2003b; Amundsen & Knudsen 2009). Habitat shift to shallower depths during winter has also been found for lake trout (Salvelinus namaycush Walbaum) in a recent telemetry study, in which the mean daily depth during the ice-covered season was explained by ambient light conditions, measured as the combination of snow depth and day length (Blanchfield et al. 2009). Arctic charr is opportunistic and responds rapidly to changes in resource availability and altered ecological interactions (Langeland & L’Abée-Lund 1998; Klemetsen et al. 2002; Gregersen et al. 2006). Therefore, the habitat shift probably is a result of seasonal changes in the competitive situation, because of interspecific variation in ability to cope with the challenging winter period. Brown trout suffer more from cold water and low light levels during the ice-covered period compared to Arctic charr, which has a higher food intake and lower energy loss. Hence, years with a late ice breakup and a long winter will cause stronger competition pressure on brown trout from Arctic charr, as shown by our findings. This is in accordance with a number of studies where inter-annual variations in seasonal climatic events, in combination with competition for resources, have been reported to influence temporal variations in relative species abundance (e.g. Berryman & Lima 2006; Grant & Grant 2008; Previtali et al. 2009).

Competition intensity can vary over time because of temporal variation in resource limitation (Schmitt & Holbrook 1986; Hoset & Steen 2007). If competing species exhibit differing competitive abilities under contrasting environmental conditions, such temporal variations may prevent competitive exclusion (Lauer & Spacie 2004; Descamps-Julien & Gonzalez 2005; Suwabe et al. 2009). The coexistence of brown trout and Arctic charr may be regulated by similar seasonal dynamics, where the more aggressive, energy demanding behaviour of brown trout is beneficial during the summer when resources are abundant, while the energetically efficient Arctic charr is the superior competitor in cold periods with less food.

Ice-cover conditions may influence both production and taxonomic composition of plants and invertebrates in aquatic systems (Douglas & Smol 1999). Furthermore, snow and ice conditions appear to have an increasingly dominant role in structuring aquatic ecosystems with increasing latitude (Sorvari, Korhola & Thompson 2002; Ruhland, Priesnitz & Smol 2003). Settling and thawing of ice are threshold processes that, when exceeded, may initiate abrupt regime shifts in aquatic community compositions, as revealed by a recent palaeoecological study on algae and invertebrate communities in Arctic lakes (Smol et al. 2005). Using northern salmonid fish species as a model, we have shown that interspecific variation in performance under ice may transfer to competitive interactions and potentially influence the geographical distribution of species through competitive exclusion. Similar variability in climate response is also expected to occur in other taxonomic groups, and complex ecosystem changes may take place as ice-cover conditions changes in a new climate.


We thank the staff at NINA Research Station at Ims for technical assistance during experiments and O. Hegge, J. Skurdal, L. Fløystad and R. Saksgård for participation in test fishing in Lake Atnsjøen. A. Foldvik is thanked for help with ArcGIS. All animal experiments were carried out according to national regulations for the treatment and welfare of experimental animals. The study was supported by the Norwegian Research Council (NFR 185109/S30) and the Norwegian Institute for Nature Research.