Statistical Analysis

Statistical analysis were all conducted using the R software (R Development Core Team, 2013). Diversity and richness estimates were computed based on raw counts of reads attributed to each OTU, using the R package ‘Vegan.’ Richness was estimated by rarefaction, as the expected number of OTUs in a random subsample of each sequence library, having the size of the smallest library (Hurlbert, 1971). Diversity was estimated by the Simpson index, calculated as (Simpson, 1949) and evenness as (Pielou, 1966) with S being the observed number of OTUs and fi the frequency of each OTUi in the sample. To evaluate overall differences between eukaryotic assemblages, pairwise Bray–Curtis dissimilarities were calculated between all samples ( with x_ji and x_ki the abundances of OTU_i in samples j and k, respectively, and S the number of OTUs observed in libraries j and k), based on OTU percentages of reads (instead of raw counts to not consider differences due to different numbers of reads). They were computed using the R ‘Vegan’ package (Oksanen et al., 2013). The same package was used to draw Non-metric MultiDimensional Scaling (NMDS) plots comparing communities from the 2-years survey of both ecosystems or on La Claye and Ru Sainte Anne separately. They were based on Bray–Curtis dissimilarities calculated after square-root transformation and Wisconsin standardization (Bray and Curtis, 1957) of OTU percentages. Ellipses were drawn on NMDS plots using the R package ‘Ade4’ (Dray and Dufour, 2007) to highlight whether the communities were collected in water or sediment. Boxplots were drawn with notches to indicate whether the medians of the represented distributions could be considered as different (Chambers et al., 1983).