2.6 Statistical analysis
The normality and homogeneity of variances of the dataset in this study had been verified. The differences between three groups regarding soil physical, chemical properties and microbial α-diversity indices had been obtained by one-way analysis of variance (ANOVA). While group means had been subjected to Duncan’s test at p <0.05 using a DPS (Data Processing System) v.7.05. The concentrations of 18 soil metals among three sampling groups were shown by heatmap, generated by R software v.3.5.1 through the heatmap package. Principal Co-ordinates Analysis (PCoA) of soil microbial community based on ASVs level, analysis of similarities (ANOSIM), and variance partitioning analysis (VPA) for microbial community composition among groups were all using the vegan package in R software v.3.5.1.
A network analysis was conducted to reveal the relationship between bacterial and fungal taxa (relative abundance > 0.1%), soil metals, and physicochemical properties. We calculated the pairwise Spearman’s correlation coefficients (p <0.05) using psych package in R software v.3.5.1 and visualized the network using Gephi software v.0.9.2. Each node represented one microbial taxon or soil parameter. Red lines and blue lines represented strong positive and negative correlations, respectively. The destruction resistance of the network was generated using R software v.3.5.1 through randomly removing the nodes and calculating the natural connectivity and network degree values.
The direct or indirect effects of total petroleum hydrocarbons on soil parameters, bacterial and fungal diversity, and community compositions were generated through conducting the Spearman correlation or Mantel test between matrices (p <0.05).