3.4 Influence of landscape on genetic diversity
Analysis of the db-RDA model in Shunchang County showed that different landscape types had significant effects on the genetic diversity at three scales (Table 4). At the three scales of 300 m, 800 m, and 1,000 m, the percentage of P. massoniana was significantly related to the genetic diversity of this species (P (300 m) = 0.031;P (800 m) = 0.022; P (1,000 m) = 0.006), explaining 47.478%, 45.730%, and 43.342% of the variation, respectively (Table 4). Urban landscape was significantly correlated with the genetic diversity only at 800 m (P = 0.046), whereas mixed forest with hosts had significant effects at scales of 800 m and 1,000 m (P (800 m) = 0.038; P (1,000 m) = 0.031). At the 1,000 m scale, additional types of landscapes had significant correlations, including P. elliottii (P = 0.04), farmland (P = 0.018), and roads (P = 0.046) (Table 4). In Xiapu, in addition to the influence of P. massoniana at the three scales, other landscape types at different scales also showed significance (Table 5). At the 800 m scale, P. elliottii was significantly correlated with the genetic diversity (P = 0.019), explaining 23.237% of the variation (Table 5). At the 1,000 m scale, similar to Shunchang, there were more relevant landscape variables than at the two smaller scales, primarily including mixed forest with hosts (P = 0.032), roads (P = 0.032), farmland (P = 0.007), and urban (P = 0.041) (Table 5). In both areas, the different landscape types could affect the genetic diversity of M. alternatus , and the number of landscape types with effects increased with the increase in scale.
GAM analysis was performed to explain the nonlinear relationship in the db-RDA analysis. According to the results, the landscape variables fit well on the first two dimensions of the PCoA ordination (Figures 5 (Shunchang) and 6 (Xiapu)). Pinus massoniana increased gradually with the first PCoA axis or the second axis in both areas, indicating that P. massoniana was positively correlated with genetic diversity (Figure 5A, B, E; Figure 6A, B, E). The landscape types that increased along the first PCoA axis also included roads in Shunchang (Figure 5I), whereas mixed forest with hosts and P. elliottiimainly increased with the second PCoA axis (Figure 5C, F, G). In Xiapu, mixed forest with hosts at the 800 m scale and P. elliottii and mixed forest with hosts at the 1,000 m scale also increased with the first PCoA axis (Figure 6C, D, F, H). Urban decreased with the second PCoA axis in both areas (Figure 5D; Figure 6G), whereas farmland decreased with the first PCoA axis or the second axis (Figure 5H or Figure 6I), which indicated that the genetic diversity was negatively correlated with urban and farmland landscapes.