TABLE 3 Multiple matrix regression with randomization (MRR) analyses on community composition (Simpson dissimilarity index, βSIM) of Quercus alnifolia (Qa) sampling sites at ASV and OTU levels. The explanatory variables were backward selected from full models containing the following distance matrices: ‘flat’ scenario (NULLIBR), topography (SPATWD), topoclimate (ENVPC1-2), topographic complexity (TRIIBR) and forest fragmentation (FRAIBR). The significance of both explanatory terms retained in the best-fit model and those rejected during the backward selection procedure are reported. The independent (R I2) and joint coefficient of determination (R J2) of each predictor retained in the best-fit final model are provided.