Competition is assumed to shape niche widths, affecting species survival and coexistence. Expectedly, high interspecific competition will reduce population niche widths, whereas high intraspecific competition will do the opposite. Here we test in situ how intra- and interspecific competition affects trophic resource use and the individual and population niche widths of two lacustrine fish species, Arctic charr and brown trout, covering a 40 year study period with highly contrasting competitive impacts prior to and following a large-scale fish culling experiment. Initially, an overcrowded Arctic charr population dominated the study system, with brown trout being nearly absent. The culling experiment reduced the littoral Arctic charr density by 80%, whereupon brown trout gradually increased its density in the system. Thus, over the study period, the Arctic charr population went from high to low intraspecific competition, followed by increasing interspecific competition with brown trout. As hypothesized, the relaxed intraspecific competition following the experimental culling reduced individual diet specialization and compressed population niche width of Arctic charr. During the initial increase of the brown trout population, there was a large dietary overlap between the two species. Over the subsequent intensified interspecific competition from the population build-up of brown trout, their trophic niche overlap chiefly declined due to a dietary shift of Arctic charr towards enhanced zooplankton consumption. Contrary to theoretical expectations, the individual and population niche widths of Arctic charr increased with intensified interspecific competition. In contrast, the diet and niche width of brown trout remained stable over time, confirming its competitive superiority. The large-scale culling experiment and associated long-term research revealed pronounced temporal dynamics in trophic niche and resource use of the inferior competitor, substantiating that intra- and interspecific competition have large and contrasting impacts on individual and population niches.

Fish sampling and processing
Charr and trout were sampled annually from the littoral habitat (<15 m depth) in August from 1980 to 2019 using single-meshed gillnets of various mesh sizes prior to 1989 and thereafter multi-meshed gillnets with panels of eight different mesh sizes ranging from 10 to 45 mm, knot to knot (Table 1). The nets fished in the lake overnight for approximately 12 hours. Fork length and other parameters not used in the current study (weight, sex, and gonad maturation) of all fish were recorded in the field and stomach samples were collected. Catch per unit effort (CPUE), defined as the number of fish caught per 100 m2 gillnets per night, was estimated as a proxy for the littoral abundance of charr and trout.

In the lab, stomachs were opened, and the fullness degree was determined on a scale from 0 to 100% (Amundsen and Sánchez‐Hernández 2019). Prey items were identified at the lowest taxonomical level, and their relative contribution to total stomach fullness (expressed in percentage) was calculated according to Amundsen (1995). Prey taxa were then grouped into twelve categories: (I) small cladoceran zooplankton (Bosmina spp.), (II) large cladoceran zooplankton (Daphnia spp. and Holopendium gibberum), (III) predatory cladoceran zooplankton (Bythotrephes longimanus, and Polyphemus pediculus), (IV) copepod zooplankton (cyclopoid and calanoid copepods), (V) amphipods (Gammarus lacustris), (VI) mollusks (Radix peregra, Planorbis sp., Valvata sp., and Pisidium sp.), (VII) pleuston (terrestrial and hatching aquatic insects), (VIII) Chironomidae pupae, (IX) Chironomidae larvae, (X) Trichoptera larvae (house-living and free-living), (XI) other benthos (Ephemeroptera nymphs, Plecoptera nymphs, Megaloptera larvae, Tipulidae larvae, Coleoptera, and the chydorid cladoceran Eurycercus lamellatus.), and (XII) fish (three-spined stickleback, charr, and unidentified fish remains). These prey categories were used for a simplified visualization of temporal dietary changes, whereas un-pooled prey data were used for the subsequent dietary analyses.

For the dietary analyses, stomachs with a fullness degree below 10% or containing only unidentified prey were removed from the dataset. Each individual stomach content was then standardized to estimate prey abundance as the mean contribution of each prey category to the diet. The fish were divided into the three size classes (<150 mm, 150-299 mm, and >300 mm) to study ontogenetic dietary shifts over the 40 year study period and if these shifts might be influenced by an increase in individual dietary specialization as explained earlier, where individuals switch to alternative resources to mitigate the effects of competition (Araújo et al. 2011). By pooling data in five-year sampling periods, the 150 and 299 mm size class provided large enough sample sizes for temporal comparisons. Since the number of samples in this size class ranged from 86 to 200 individuals for charr and 60 to 129 for trout, 86 charr and 60 trout stomachs were randomly selected from each sampling period to avoid sample size bias in subsequent analyses. No significant size differences among sampling periods (ANOVA, all P>0.05) were detected within each size group. A total of 1424 charr and 621 trout were included in the analyses.

Statistical analysis Descriptive and inferential analyses were performed with the open-source software Rstudio (version 1.1.423, Rstudio Inc.), based in R (version 3.5.1, R Core Team). We used a permutational multivariate analysis of variance (PERMANOVA) to assess dietary composition differences between sampling periods and host species (Anderson 2005). A Bray-Curtis based non-metric multidimensional scaling (NMDS) was further used to graphically illustrate any dietary differences between charr and trout among different sampling periods. To determine which prey contributed the most to the observed differences, we opted for Sum-of-LR, a multivariate method based on generalized linear model with negative binomial errors (Warton et al. 2012, Wang et al. 2012). We chose this method over the more widely used similarity percentage (SIMPER) analysis as the latter can confound strong between-group effect with large within-group variance, yielding misleading results (Warton et al. 2012).

We measured the total niche width (TNW) of populations applying the Shannon index of diversity to the population’s distribution of resource use (Roughgarden 1979). We then partitioned TNW into the within-individual component of niche width (WIC), which is the average individual niche width, and the between-individual component of niche width (BIC), which is the variation between individuals’ niche positions, such that TNW = WIC + BIC (Roughgarden 1972). To assess the impact of trout density on charr’s niche, we correlated TNW, WIC, and BIC values with CPUE using Spearman’s correlation coefficient with Bonferroni’s correction. To evaluate the degree of individual diet specialization, we used multiple measures for a more robust assessment of this multifaceted trait than can be accomplished using a single metric. Individual diet specialization can be expressed as the variation between an individual diet and the population diet or between an individual and other individuals. We therefore calculated the WIC/TNW ratio, which provides a measure of specialization by individuals within a population, with specialization being high when WIC/TNW is low. Additionally, the degree of individual diet specialization was assessed with the level of diet variation (E; Araújo et al. 2008), the proportional similarity index (PS_i; Bolnick et al. 2003), and the individual specialization index (IS and V; Bolnick et al. 2002, 2007).

We used variance inflation factor (VIF) to detect multicollinearity (correlation between predictors) among individual specialization indexes. A VIF value < 3 indicate lack of collinearity (Zuur et al. 2010). Collinearity was detected among indexes (all VIF values >3); hence, we opted to use WIC/TNW values to represent individual specialization. Finally, we tested relationships between sampling periods, WIC, BIC and WIC/TNW values using a generalized least squares model (GLS) using the nlme package (Pinheiro et al. 2021). To account for temporal autocorrelation, we used the autoregressive term AR1. Model fit was evaluated with the autocorrelation function ACF and partial autocorrelation function PACF and the fit between residuals versus fitted values. Data from all sampling periods were used to assess the first two hypotheses with the intent of inferring if temporal changes in individual and population niche widths in charr were likely due to decreased intra-specific competition or increased inter-specific competition. To assess the impact of trout density on charr’s individual specialization, we correlated WIC/TNW values with CPUE using Spearman’s correlation coefficient with Bonferroni’s correction. Calculation of TNW, WIC, BIC, WIC/TNW, E, PS_i and IS were performed with the RInSp package (Zaccarelli et al. 2013).

Interspecific diet overlap was calculated with the Schoener’s overlap index α=1-  1/(2 )(Σ∣Pxj-Pyj∣ x 100 (Schoener 1970), where Pxj and Pyj are the relative abundance of diet item j in the stomach of species x and y, respectively. The index ranges from 0 to 100% with values of 0 indicating absence of diet overlap and values of 100% indicating a complete dietary overlap. Additionally, for the 150-299 mm size classes of charr and trout, we calculated the pairwise diet similarity (PSij) between each pair of heterospecific individuals i and j: (PS)ij=∑k(min⁡(Pik ),Pjk), where Pik and Pjk are the proportions of the Kth prey type in individual i’s and j’s diet (Bolnick and Paull 2009). A value of 0 indicates that the paired individuals do not share common prey, while values close to 1 indicate that they consume the same prey in identical proportions. (PS)ij was calculated with the RInSp package (Zaccarelli et al. 2013)

Effects of resource pulses We run additional analyses in order to test whether the outcomes remain the same after excluding a resource pulse, i.e. infrequent, large‐magnitude and short‐duration events of increased resource availability (Yang et al. 2010). Temporally superabundant food sources might lead to a convergence in the resource use of co-occurring predators, altering their immediate trophic interactions (Lack 1946, Croxall et al. 1999, Selva et al. 2012). More specifically, the superabundance of a single prey potentially may temporally influence individual specialization and resource partitioning (Meyer, 1989; Malmquist et al., 1992; Robinson and Wilson, 1998). In subarctic lakes, hatching chironomid pupae cyclically occur in superabundance during midsummer, constituting a resource that is typically included in the fish diet when abundantly present (Adalsteinsson, 1979; Amundsen and Klemetsen, 1988). A superabundance of Chironomidae hatching and emergence were observed in the field within several of the sampled years (1980, 1986, 1994, 2002, 2007, 2011, 2014, and 2018). This massive hatching is mainly by a single species, Heterotrissocladius subpilosus, and lasts for only 2-3 weeks in early summer. The species strongly dominates the profundal benthos as larvae (Klemetsen et al. 1992). Given a particularly strong presence of Chironomidae pupae in 1980, which was the only observation available for the pre-culling period, we also addressed our research hypotheses following the exclusion of this prey type. We excluded Chironomidae pupae to reduce bias in interspecific competition metrics among sampling periods as events of Chironomidae pupae superabundance would have been more diluted among pooled periods compared to a single event. Hence, we repeated the above procedures and analysis on a subset of 464 charr and 294 trout excluding this prey from the diet.