Statistical Tests

A maximum likelihood-based phylogenetic tree for analyzing NRI was calculated using MEGA 5.0 (Tamura et al., 2011) using nearest neighbor interchange. Calculations for NRI are detailed in Webb (2000) and Horner-Devine and Bohannan (2006).

Correlations between bacterioplankton community composition (Bray–Curtis distances) and environmental factors versus spatial factors (Euclidean distances of salinity, temperature, Chl a, ammonium [NH₄⁺], nitrate [NO₃^-], phosphate [PO₄^-], and silicate [SiO₄] compared to latitude and longitude) were calculated using standard Mantel’s tests for environmental [E] and spatial [S] factors and partial Mantel’s tests for the fraction of bacterioplankton community composition that can be explained by the environmental factors independent of any spatial structure [E| S] and the fraction that can be explained by spatial allocation independently of any environmental variables [S| E] (Table ​Table11). Assignment to the most appropriate metacommunity type was performed following Cottenie (2005).

Summary of Mantel’s tests performed in the present study.

Niche breadth was calculated using Levin’s niche breadth index following Pandit et al. (2009) (B = 1/Σ^N_i = 1p²_ij) where p_ij is the proportion of OTU j in the sample i and N is the number of samples. OTUs with high B values are classified as habitat generalists and evenly distributed along a wide range of habitats. In contrast, OTUs with low B values are considered habitat specialists and unevenly distributed among sampling sites. Mantel’s tests for [E] and [S] and partial Mantel’s tests for [E| S] and [S| E] was performed for these different niche breadth groups as above.

All statistical tests and graphical outputs were performed in R 3.2.2 (R Development Core Team, 2014), using the packages Vegan (Oksanen et al., 2010), picante (Kembel et al., 2010), and ggplot2 (Wickham, 2009).