1. Novel trophic interactions between invasive and native species potentially increase levels of inter-specific competition in the receiving environment. However, theory on the trophic impacts of invasive fauna on native competitors is ambiguous, as while increased inter-specific competition can result in the species having constricted and diverged trophic niches, the species might instead increase their niche sizes, especially in omnivorous species.

2. The competitive interactions between an omnivorous invasive fish, common carp Cyprinus carpio, and a tropically analogous native and threatened fish, crucian carp Carassius carassius, were tested using comparative functional responses (CFRs). A natural pond experiment then presented the species in allopatry and sympatry, determining the changes in their trophic (isotopic) niche sizes and positions over four years. These predictive approaches were complemented by assessing their trophic relationships in wild populations.

3. CFRs revealed that compared to crucian carp, carp had a significantly higher maximum consumption rate. Coupled with a previous cohabitation growth study, these results predicted that competition between the species is asymmetric, with carp the superior competitor.

4. The pond experiment used stable isotope metrics to quantify shifts in the trophic (isotopic) niche sizes of the fishes. In allopatry, the isotopic niches of the two species were similar sized and diverged. Conversely, in sympatry, carp isotopic niches were always considerably larger than those of crucian carp and were strongly partitioned. Sympatric crucian carp had larger isotopic niches than allopatric conspecifics, a likely response to asymmetric competition from carp. However, carp isotopic niches were also larger in sympatry than allopatry. In the wild populations, the carp isotopic niches were always larger than crucian carp niches, and were highly divergent.

5. The superior competitive abilities of carp predicted in aquaria experiments were considered to be a process involved in sympatric crucian carp having larger isotopic niches than in allopatry. However, as sympatric carp also had larger niches than in allopatry, this suggests other ecological processes were also likely to be involved, such as those relating to fish prey resources. These results highlight the inherent complexity in determining how omnivorous invasive species integrate into food-webs and alter their structure.

Comparative functional responses

The crucian carp used in the CFRs were captured using baited traps from natural populations in two adjacent ponds in Southern England. Both populations had been seeded from the source and so did not differ genetically, with two ponds used to ensure the appropriate number of fish were collected. The carp were sourced from a local hatchery where they had been pond-reared on a mix of natural and supplemental food, before being held in small outdoor ponds (1000 L) for 2 months without supplemental feeding to promote natural foraging behaviours. Both species were then moved into an aquarium facility and held in species-specific holding tanks (90 L; 10 fish per tank) at 17 °C for 21 days, with daily feeding (ad libitum) with frozen chironomid larvae. The CFR experiment was completed at 17 °C to represent typical summer water temperatures in England (Britton, 2007). Given the two species are aggregative in nature (Penne & Pierce, 2008; Baumgartner et al., 2008; Bajer et al., 2011), then rather than complete the experiment on individuals, the fish were used in conspecific pairings. Prior to the CFR trials, the fish were measured with callipers (standard length (SL) to 0.1 mm), with mean lengths (± SD) of crucian carp being 66.3 ± 6.7 mm and common carp being 66.6 ± 7.7 mm, and were thus considered as size matched.

The CFR trials were all completed in rectangular tanks (20 L volume) without substrate or refugia, and to eliminate external stimuli, were covered with a lid and the sides were also covered. Hunger levels were standardized by the experimental fish not being fed for 24 h prior to experiments. The paired conspecifics were selected randomly from their holding tanks, released into the experimental tanks and then acclimated for two hours. The food resource was pelletized fishmeal (‘pellets’) of 2 mm diameter, as these provide a resource of standard dimensions that have been consumed readily by similar fish species in functional response experiments (e.g. Murray et al., 2013). The pellets were released into the tanks at one of seven specific amounts (2, 4, 8, 16, 32, 48 and 96 pellets), with each amount replicated at least three times. Each individual trial lasted four hours and, at their conclusion, the fish were removed from the tank and the number of unconsumed pellets counted. The derived number of consumed pellets thus represents the number consumed per conspecific pair, rather than per individual fish.

Following the conclusion of all trials, the CFRs were modelled in the R package “frair” (Pritchard, 2014) using maximum likelihood estimation (MLE; Bolker, 2010) and Rogers’ (1972) Random Predator Equation (Equation 1), as the prey were not being replaced as they were consumed. Where the proportion of prey consumed decreased as prey density increased then the logistic regression produces a significantly negative result representing a ‘Type II response’; in contrast, if it produces a significantly positive result then it represents a ‘Type III’ response (Juliano, 2001). Given both species indicated a significant Type II response, then Rogers’ random predator equation was determined from:

N_e = N ₀(1 – exp (a(N _eh – T)))

where N_e  is the number of pellets eaten, N ₀ is the initial density of pellets, a  is the attack parameter, h  is the handling parameter and T  is the total time available (fixed at 1). The FR data were non-parametrically bootstrapped (bias corrected and accelerated) (n = 2000) to generate 95% confidence intervals around the mean FR curve of each species, with comparison of the 95% confidence intervals enabling these data to be considered in a phenomenological manner with regards to population level inferences (Pritchard et al., 2017) and with overlapping confidence intervals considered as indicating non-significant differences in the FR curves of the two species. In addition, parameter estimates [a, h] between the two species were compared using the z-method (Juliano, 2001) via frair:fair_compare.

Natural pond experiment

Predicting the trophic interactions of crucian carp and carp was completed in a natural pond experiment completed in southern England between 2016 and 2019. In January 2016, juveniles of both species were sourced from local hatcheries where they had been reared in ponds, and were released into three adjacent (but unconnected), fishless (following their draining, drying and re-filling), former aquaculture ponds of approximately 400 m^-2, maximum depths of 1.2 m, and with relatively clear water (secchi disk depths > 0.75 m) and highly abundant macrophyte growth (mainly Elodea spp.), with the water clarity remaining largely unchanged throughout the experimental period. The ponds were used as three distinct treatments, but with these not replicated due to logistical reasons preventing use of a greater number of ponds. Two of the ponds were used as allopatric controls, with 100 juvenile carp released into one pond and 100 crucian carp into the other (all < 100 mm). The third pond was used as a sympatric treatment, where the same number of fish were used (100) but split 50:50 between both species. As both species lack external features to enable differentiation of the sexes, then the sex ratios were unknown. While all three ponds had an increasing number of invasive signal crayfish Pacifastacus leniusculus present during the study (‘crayfish’ hereafter), only in the sympatric treatment in 2018 and 2019 were sample sizes sufficiently high to enable samples to be analysed (n ≥ 6).

The ponds were then left until September 2017 to enable their tissues to become isotopically equilibrated to their new prey resources. The fish were then sampled in September 2017, 2018 and 2019 using baited fish traps set overnight. After lifting, the captured fish were removed, measured (fork length, FL, nearest mm), anaesthetized and a fin biopsy taken, and were then released back into their pond. Concomitantly, samples of macro-invertebrates (as fish putative prey resources) were taken using a sweep net and sorted for stable isotope analysis (SIA). The samples of fish fin, crayfish and macroinvertebrates were then taken to the laboratory, dried to constant mass at 60 ^°C and then analysed at the Cornell University Stable Isotope Laboratory (New York, USA) for δ¹³C and δ¹⁵N in a Thermo Delta V isotope ratio mass spectrometer (Thermo Scientific, USA) interfaced to a NC2500 elemental analyser (CE Elantach Inc., USA). Analytical precision of the δ¹³C and δ¹⁵N sample runs was estimated against an internal standard sample of animal (deer) material every 10 samples, with the overall standard deviation estimated at 0.08 and 0.04 ‰ respectively. Ratios of C:N were generally between 3.5 and 4.0, and so were not mathematically corrected for lipid (Winter et al., 2021).

Prior to further analyses, the SI data of the fish putative prey were compared within each pond by year and between the ponds. As these revealed some considerable differences (Supplementary material; Table S1) then the fish SI data could not be compared directly between the ponds and years without correction (De Santis et al., 2021). Consequently, the δ¹⁵N muscle data were converted to trophic position (TP) according to (Olsson et al., 2009):

TP = 2 + δ¹⁵N_fish - δ¹⁵N_prey / 3.4

where TP and δ¹⁵N_fish are the trophic positions and the nitrogen ratios of each individual fish, δ¹⁵N_prey is the mean nitrogen ratio of the putative macroinvertebrate prey resources (Table S1), 2 is the trophic position of these prey resources (as primary consumers) and 3.4 is the generally accepted fractionation factor between adjacent trophic levels (Post, 2002). The fish δ¹³C data were converted to corrected carbon (δ¹³C_corr) according to the following equation (Olsson et al., 2009):

δ¹³C_corr = (δ¹³C_fish - δ¹³C_meanMI) / CR_MI

wherein δ¹³C_fish is the δ¹³C value of each fish, δ¹³C_meanMI is the mean δ¹³C of the macroinvertebrate prey (Table S1) and CR_MI is the carbon range (δ¹³C_max - δ¹³C_min) of the same macroinvertebrates (Olsson et al., 2009).

Following the correction of the SI data to δ¹³C_corr and TP, the initial data analysis tested differences in these data between the two fish species in the sympatric treatment using ANCOVA, where the covariate was fish length and data for all years were combined. Then, to account for the presence of crayfish in the sympatric treatment, the significance of differences between their corrected SI data with the fish corrected SI data were tested in one-way ANOVA (with Tukey multiple comparisons of means with 95 % family-wise confidence levels). The corrected SI data were then used to calculate the trophic niche size of each fish species per pond and sampling year, using the isotopic niche as a proxy of the trophic niche (Jackson et al., 2011). Whilst closely related to the trophic niche, the isotopic niche is also influenced by factors including growth rate and metabolism (Jackson et al., 2011). The isotopic niches were calculated as standard ellipse areas (SEA) in SIBER (Jackson et al., 2011; Jackson et al., 2012). SEAs are a bivariate measure of the distribution of individuals in isotopic space and as the ellipses enclose the core 40 % of data, they represent the typical resource use of the analysed population (Jackson et al., 2011; De Santis et al., 2021). A Bayesian estimate of SEA (SEA_B) tested differences in niche sizes between the treatments per species, calculated using a Markov chain Monte Carlo simulation (10⁴ iterations per group) (Jackson et al., 2011; Jackson et al., 2012). Differences in the size of isotopic niches (as SEA_B) were evaluated by calculating the probability that the relative posterior distributions of the niche size of the allopatric treatment were significantly smaller or larger than those of each of their sympatric niches (a = 0.05) in SIBER. The SI data were then used to calculate isotopic niche overlap (%) between the species using SEA_c also calculated in SIBER, where subscript ‘_c’ indicates a small sample size correction was used (Jackson et al., 2012). Use of SEA_c was mainly to get a representation of the extent of niche overlap between species, as it is more strongly affected by small sample sizes (< 30) than SEA_B (Jackson et al., 2012). Overlaps between the isotopic niches were calculated based on SEA_c with 95% confidence tested for the species between their allopatric and sympatric treatments each year.

Wild ponds with sympatric carp and crucian carp

There were four wild ponds sampled for their populations of sympatric carp and crucian carp between July and September 2019. The ponds were all located in southern England, were between 0.5 and 1.5 ha in area and had depths to 2 m. Their exact locations are unable to be provided to protect business confidentiality, as each was run as a private fishery for catch-and-release angling. All of the fish had been present in the ponds for at least three years (i.e. there had been no recent stocking of fish). The fish were sampled by a combination of baited fish traps and rod and line angling during stock assessment exercises, where the species were identified, measured (FL, nearest mm) and scale samples taken (3 to 5 scales per fish), originally for age and growth analyses for fishery management purposes. It was these scales that were used for stable isotope analysis, with scales tending to have a longer isotopic half-life than fin tissue (Busst & Britton, 2018). The scales were not decalcified prior to isotopic analysis, as the removal of inorganic carbonates has no significant effect on scale δ¹³C and δ¹⁵N values (Ventura & Jeppesen, 2010; Woodcock & Walther, 2014). They were prepared by their cleaning with distilled water before the outer portion of the scale was removed for SIA, as this ensures that the analysed tissue is from the most recent growth of each fish (e.g., the last full year of growth; Hutchinson & Trueman, 2006). Only one scale was analysed per individual fish as this provided sufficient material for SIA. The samples were then prepared and analysed for δ¹³C and δ¹⁵N as per the natural pond experiment. As these SI data were only compared between the two species within each pond and not between ponds, no corrections were made to these data. As per the natural pond experiment, differences in the SI data between the species were initially tested in ANCOVA before their isotopic niches were calculated (as SEA_B and SEA_c) in SIBER.