Figure 2: Non-metric multidimensional scaling (NMDS) of dietary samples from (a, d) the pilot dataset (n = 75), (b, e) the same samples sequenced during the pilot, but from the full sequencing run (n = 75) and (c, f) the full sequencing with all samples (n = 437). Ordinations in (a, b, c) are based on Bray-Curtis dissimilarities, whereas (d, e, f) are based on Jaccard-indexed dissimilarities computed from presence-absence data (zOTU presence = >0.01%). Colors distinguish the season of sampling, whereas shapes denote the identity of the consumer. Overlayed ellipses indicate 95% confidence levels, and accompanying centroids show the average within-group positions. Stress and number of dimensions (demanding a stress of ~0.1 or less) are shown. Significant differences between levels of groups (seasons, stations, and species) are shown based on PERMANOVAs, and in bold if the group dispersion was found to be homogenous (betadisper).