Spatial patterns in mussel richness and endemism (i.e., weighted and corrected weighted endemism) at stream reach, HUC4, HUC6, and HUC8 spatial scales were characterized. Stream SR was estimated by summing the total number of mussel species predicted within each stream. Stream-level weighted endemism was calculated by weighting each species’ presence in the stream by the inverse of the total number of streams it was predicted to occupy, then summing across all species predicted present within the stream. This biodiversity metric therefore provides context about the range extent of the species predicted within the stream (e.g., higher values indicate either high richness or the presence of several range-restricted species). Stream-level corrected weighted endemism was calculated by dividing weighted endemism by the total number of species predicted present within the stream. This metric therefore corrects for an abundance of species predicted present within a stream (meaning, larger corrected weighted endemism values indicate the presence of more range-restricted species). Endemism metrics were rescaled to fall between zero and one to increase metric interpretability. These rescaled values were used in ensuing analyses.
Stream-level SR and endemism metrics within each HUC region were strongly right-skewed (many smaller values and few large values). This necessitated calculating the median value of these richness and endemism metrics across all streams within each HUC region. Next, local neighborhoods for each HUC4, HUC6 and HUC8 region were generated to account for the degree of spatial connectivity among HUC regions (i.e. the number of neighboring regions for each HUC). These neighborhoods, as well as HUC median richness and endemism values, were then used to estimate 1) a global Moran’s I across the entire study region and 2) a local Getis–Ord Gi statistic for each HUC region. The global Moran’s I analysis evaluates directionality in, and magnitude of, spatial autocorrelation (i.e., whether high or low values cluster together, are randomly distributed or are overdispersed) throughout an entire region. Alternatively, the local Getis–Ord Gi statistic identifies clusters of zones within the entire region that contain statistically greater or lesser values of interest. HUC regions associated with statistically significant (α = 0.05, P < 0.025 and P > 0.975) values from the Getis-Ord analysis were classified as hot or cold spots of mussel richness or endemism.
