Estimation of Uncertainty and Location-specific Variation in Regional Food Web Interactions

We demonstrate a methodology for estimating uncertainties and intra-regional variations in bird-plant food web interactions, based on transect observations at multiple locations with the region. We model species-species interaction measurements at each location as Poisson random variables whose mean is proportional to the interaction strength at that location. We also model the location-dependent interaction strength as having uniform probability on an interval. A Bayesian approach is used to estimate this interaction strength interval based on measurements at specific locations. We present and discuss the implications of graphical representations to show how the estimated means and ranges of regional interaction strength (and the variances of these estimates) depend on measurements, in the case where measurements are taken at 3 locations. Among other properties, we show that taking the mean of three measured interaction frequencies will tend to underestimate the actual mean interaction strength, assuming a uniform interval model. The methodology is applied to a specific case of transect survey data for avian frugivore species interactions with fruit-bearing plants from three locations within each of six regions in Ecuador. We hypothesize that species within the same genus tend to substitute for each other out in different locations within the same region. We test this hypothesis by comparing the estimated location-dependent variances of species-species and genus-genus interactions. Results contraindicate the hypothesis: different species from the same genus tend to have higher interaction strengths in similar locations, i.e. "species in the same genus flock together." We also find some evidence that fragmentation of the environment due to human activity, even if it does not reduce species diversity, may have a uniformizing effect so that the site-to-site variability in bird-fruit interactions is reduced.