TABLE 2 Best-fit models from distance-based redundancy analyses (dbRDA) on community composition (Simpson dissimilarity index, βSIM) of sampling sites at ASV and OTU levels, performed either with all sites (across habitats) or separately for each of the forest habitat types. The explanatory variables were forward selected from full models containing three sets of predictors: forest habitat type (HAB), spatial (SPAPCNMi) and topoclimatic (ENVPCi) variables. The adjusted coefficient of determination (R 2ADJ) for each model is provided. The Pinus nigra (Pn) and Juniperus foetidissima (Jn) sampling sites were both separately and jointly analyzed according to results of NMDS-based ordinations (Figure 2). Predictor information for models with no significant variables (null) is replaced by dashes.