Abundances of bacterial OTUs, families, orders, classes and phyla were individually modelled using NBMs. Sample sizes are N=306 for leaves and N=310 for roots. Significance was assessed using a Likelihood ratio test at P<0.05 after correction for multiple comparisons using Benjamini–Hochberg false discovery rate. ‘Geno’=genotype; ‘GxS’=genotype-by-site interactions; ‘AxS’=age-by-site interactions; ‘YxS’=year-by-site interactions. (a) Bar plots show the total relative abundance of bacterial taxa predicted by each source of variation (top horizontal axis) at multiple taxonomic levels (bottom horizontal axis), in variance-stabilized NBMs. (b) Effect sizes (fold changes between experimental groups for each term on the top horizontal axis) are plotted for all statistically significant pairwise contrasts predicted by variance-stabilized NBMs (Wald test, P<0.05). The bottom and top edges of the boxes mark the 25th and 75th percentiles (that is, first and third quartiles). The horizontal line within the box denotes the median. Whiskers mark the range of the data excluding outliers (green or grey dots) that fell more than 1.5 times the interquartile range below the first quartile or above the third quartile. To improve readability of the plot, the vertical axis was truncated at 211, concealing seven extreme outliers: all were root-associated OTUs with extreme changes in abundance due to age-by-site interactions.