Landscape community connectivity
When analyzing the entire sampling at Nevado de Toluca (19 km), pairwise similarity within communities from the haplotype to higher CLs (0.5 to 7.5% lineages) decreased with Euclidean (“flat”) distance (Figure 5ac). The fit of IBD was higher in Collembola (r 2 = 0.704, b = -2.07, p< 0.001 at the haplotype level, r 2 = 0.599, b = -1.13, p < 0.001 at GMYC, tor 2 = 0.580, b = -0.92, p< 0.001 at CL 7.5%; Table S3; Figure 5a) than Diptera (r 2 = 0.293, b = -0.56, p< 0.001 at the haplotype level, r 2 = 0.195, b = -0.38, p < 0.001 at GMYC, tor 2 = 0.036, b = -0.14, p< 0.001 at CL 7.5%; Table S4; Figure 5c). However, the explanatory power was slightly higher when considering effective distances by IBR (Figure 5bd; Figure S5). For Collembola, the resistance surface “Altitude 3,000” had the highest explanatory power among CLs (from r 2 = 0.723, b = -1.63, p< 0.001 at the haplotype level, r 2 = 0.644, b = -0.87, p < 0.001 for GMYC andr 2 = 0.615, b = -0.71, p< 0.001 at CL 7.5%; Table S3; Figure 5b). In Diptera the highest explanatory power was given by the resistance surface “Altitude B” (from r 2 = 0.286, b = -0.15,p < 0.001 at the haplotype level,r 2 = 0.228, b = -0.11, p< 0.001 for GMYC to 0.070, b = -0.05, p< 0.001 at CL 7.5%; Table S4, Figure 5d).
Performing DDRs analyses at even finer geographic distances in Nevado de Toluca (i.e., independently analysing communities within western or eastern sampling points), DDRs remained strong in Collembola, but were weaker or non-significant in Diptera (Figure 6). The subset of sites within the East sampling points (<5 km), also showed distance decay of similarity in Collembola (from haplotype levelr 2 = 0.485, b = -1.22, p< 0.01 to 7.5% CL r 2 = 0.141,b = -.27, p < 0.01; Table S5, Figure 6a), and Diptera, but coefficients were lower in the second (from haplotype levelr 2 = 0.110, b = -0.17, p< 0.01 to 7.5% CL r 2 = 0.031,b = -0.12, p < 0.01; Table S5, Figure 6c). When considering altitudinal limitations to dispersal with the IBR analysis, very similar results were found (Table S5, Figure S6ac). For the sites within the West section of our sampling (<2 km), a similar pattern was found for Collembola (from haplotype levelr 2 = 0.544, b = -1.36, p< 0.01 to 7.5% CL r 2 = 0.240,b = -0.41, p < 0.01; Table S5, Figure 6b) but in Diptera the tests yielded non-significant results both for the IBD and IBR analyses (Table S5, Figure 6d, Figure S6d). The slopes of the exponential decay curves in Collembola were higher than in Diptera, scales were very similar at all threshold levels, and all similarity of assemblage increased with each multi-hierarchical level (Figure 5,6; Table S3, S4). The levels of initial genetic similarity showed a significant log–log correlation with the number of lineages, initial similarity and mean similarity of communities for both our entire sampling area and focusing on the West and East sides of it (Table S6). Thus we found that community variation across genetic similarity levels can be described by a fractal geometry in each group in all the geographic scales sampled. The log-log linear correlations suggest that the patterns of assemblage variation across hierarchical levels can be described by a fractal geometry (Baselga et al., 2013, 2015).