Community analyses
Mantel test and Procrustes analyses revealed that the inter-sample community similarities between metabarcoding treatments were greatly correlated (Fig. 5; Table 2). For all cases, the Mantel correlation (i.e. Mantel R) between HTS 10 and HTS 0.5 was higher than 0.851 and P < 0.001 (Fig. 5). Among intra-pipeline treatments (HTS 10vs . HTS 0.5), 95% OTUs exhibited highest correlation (Mantel R = 0.969; Table 2; Fig. 5). Procrustes correlations, however, revealed highest values between ESVs treatments (HTS 10 vs . HTS 0.5; Procrustes correlation = 0.969, P < 0.001; Table 1), but a much lower value between 95% OTUs treatments (Procrustes correlation = 0.688; Table 1). The high community similarity among the metabarcoding data sets was also demonstrated by PERMANOVA analyses using treatment as fixed variable (P > 0.999 for all cases). Moreover, no group-specific indicator OTUs were assigned to neither HTS 10 nor HTS 0.5 treatments by indicator species analysis for any of the metabarcoding data sets.
The inter-sample community similarities (Mantel correlations) and Procrustes analyses between microscopy and HTS data also demonstrated highly correlated patterns (microscopy vs . all HTS treatments: Mantel R ≥ 0.800, P < 0.001; Procrustes correlations ≥ 0.681, P ≤ 0.001; Fig. 6, Table 1). Compared with other treatments, the 95% OTUs data set demonstrated slightly higher Mantel correlations with microscopy data (Table 1), whereas the highest Procrustes correlations were found for the ESVs HTS 0.5 data set (Table 1). Based on Random Forest analysis, the diatom assemblages in all treatments were most strongly affected by conductivity, water depth, Si, Ca, Sr, Mn and Fe, however, with different orders in variable importance (Fig. S5). Marginal tests (the significance of an individual variable when considered alone and ignoring all other variables) showed consistent patterns for the most important variables for all treatments (Table S4). The highly correlated community structures were also demonstrated in the ordination plots (Fig. 7).