3.1 | Community-wide trophic structure metrics
We assessed trophic structure change between the historical and contemporary San Juan River fish community using five community-wide metrics derived to quantify and test ecologically relevant hypotheses related to food webs (Layman et al ., 2007; Table 2). These metrics included: δ13C range (CR), δ15N range (NR), centroid distance (CD), nearest neighbor distance (NND), and standard deviation of the nearest neighbor distance (SDNND). Ranges of δ13C and δ15N are simple metrics that measure the diversity of assimilated basal resources and food chain length, respectively. Centroid distance quantifies a food web’s overall resource use diversity and is calculated as the Euclidian distance from each species’ mean δ13C and δ15N to the global centroid (i.e., the mean δ13C and δ15N value for all species). Nearest neighbor distance (NND) is a measure of the trophic similarity among species and is the mean of the shortest Euclidian distance between each species’ δ13C and δ15N mean. A food web with a smaller NND suggests trophic ecologies are similar and redundant among species while a larger NND indicates more divergent trophic niches. We also calculated the standard deviation of the NND (SDNND) because it is less influenced by sample size than NND. A comparably lower SDNND suggests differences in species’ trophic niches in a food web are more evenly distributed.
To assess differences in the five community-wide trophic structure metrics between the two periods, we used the R package SIBER(Jackson et al ., 2011) which employs a Bayesian-based Markov Chain Monte Carlo (MCMC) algorithm to produce posterior probability distributions for each metric of interest using observed δ13C and δ15N values. To produce these distributions, we used two chains of 4,000 iterations, discarded the first 1,000 iterations, included vague priors (a normal distribution for the mean and an inverse Wishart distribution for sigma), and tested for model convergence using the Gelman-Rubin ratio (R-hat) using the R package coda (Gelman and Ruben, 1992; Jackson et al ., 2011). All R-hat were <1.01. For each metric, we then calculated the probability the difference between historical and contemporary period trophic metric was greater than zero by subtracting posterior probability distributions and quantified the percent of the resultant distribution that was greater than zero. To isolate effects of species turnover on changes to the trophic structure of the community, we analyzed the community with (all species sampled) and without species turnover (only the core species present in both time periods).