Statistical analysis

The effects of land use and soil physicochemical properties on the bacterial and fungal community composition were evaluated using PAST [59] and R version 3.2.2 [60], using “vegan” package [61]. The bacterial and fungal OTU richness and Shannon diversity were calculated using the PAST function “diversity indices”. The differences in bacterial and fungal OTU richness and diversity among the five land uses were analyzed for differences among means (P < 0.05) by performing one-way analysis of variance (ANOVA) using the PAST program. All data sets were tested for normality and the equality of group variances using a JB test and the Levene statistic. One-way analysis of similarity (ANOSIM) was employed by taking the land use as the main factor to investigate the shifts in the bacterial and fungal community composition across the land uses. ANOSIM was carried out based on the Bray-Curtis distance measure using abundance data to test for significant differences in bacterial and fungal community composition among different land uses. The statistical significances of the differences in both bacterial and fungal community compositions were carried out based on 999 permutations and Bonferroni-corrected P values were applied. ANOSIM produces a sample statistic (R) that is supposed to vary between the test groups ranging from − 1 to 1 (R = 0, no separation; R = 0.30 − 0.75, there is separation but with some degree of overlapping; R = 1, complete separation) [49]. The ANOSIM produces an overall similarity measure and “Pairwise Tests” of bacterial and fungal OTUs compositions between the land uses.

Non-metric multidimensional scaling (NMDS) was performed to visualize the separation of the bacterial and fungal community compositions between each land use using R. Goodness-of-fit statistics (R²) for environmental factors fitted to the NMDS ordinations of bacterial and fungal communities were calculated using the “envfit” function in the vegan module in R, with P values based on 999 permutations. The goodness-of-fit statistics (R²) provide information about which environmental variables correspond with bacterial and fungal OTU compositions among the land uses. Autocorrelations among soil factors were investigated using Spearman’s rank correlation [62]. Land use-averaged rarefaction curves for bacteria and fungi were generated using ‘sample rarefaction’ in PAST based on seven replicates.