3.2 | Core species niche overlap and dispersion
For species present before and after river regulation (core species), we calculated species pairwise trophic overlap (i.e., niche overlap) to quantify potential changes in resource competition. We used the R statistical package nicheROVER which employs a Bayesian framework to estimate the directional probability that a randomly drawn individual of one species overlaps into the niche of another in δ13C and δ15N space (Swansonet al ., 2015; Lysy et al ., 2021). We used a resampling routine of 10,000 draws to randomly sample from each species pair, maintained nicheROVER ’s default priors (flat Normal-Independent-Inverse-Wishart distribution), and assessed model convergence (all R-hat <1.01). From each species’ directional niche overlap posterior probability distribution, we then calculated the probability the difference between time periods was greater than zero.
To identify potential changes in the diversity of trophic resources used by individual core species (i.e., collected in both time periods), we employed the SIBER package which uses a Bayesian method to calculate standard ellipses areas corrected for small sample sizes (SEAB) from observed δ13C and δ15N (Jackson et al ., 2011). We calculated SEAB because it can be interpreted as a species’ core isotopic niche breadth and because ellipses contain 40% of the sample, the metric can be used when samples sizes differ between populations (Jackson et al ., 2011; Layman et al ., 2012). Although SEAB corrects for small samples, samples size < 10 are prone to underestimates (Jackson et al ., 2011). Therefore, we included historical Colorado Pikeminnow (n = 6) in our analysis with the understanding that the SEABestimates could be artificially low. We incorporated the same MCMC algorithm parameters used to calculate fish community-wide trophic structure metrics and assessed model convergence (all R-hat <1.01). To assess whether niche dispersion differed for each core species, we calculated the probability the difference between historical and contemporary SEAB posterior probability distributions were greater than zero using the same procedure described above.