Correlation of biotic and abiotic factors to insect community composition
The ordination of FIT plots in the beetle species community applying the NMDS method yielded a significant two convergent solution (stress value = 0.04) after 20 iterations through a Bray-Curtis distance measure and clearly separated the groups according to their respective elevation (Fig. 3). The plots within the tropical Bubeng category of low elevation were slightly scattered. However, the plots in the middle and higher elevation categories showed an obviously clustered pattern. Plots from different elevation classes could be very dissimilar and the ordination ellipses for each elevation class did not overlap much. Therefore, the beetle species composition of the different elevations does not converge to the same scores (Fig. 3).
The results of a PERMANOVA model showed that major variations were distributed across the three most important environmental explanatory variables, including ELE, MTCM, and AMT, which together explained 45% of the variation in beetle species composition space.
The Mantel tests showed that β-diversity of beetles was significantly associated with tree species composition at all sampling plots at the macro scale (Table S1.2). Variation partitioning of RDA revealed that selected tree species and tree phylogenetic β-diversity separately explained 64% and 62% of the variation, respectively, with the joint effect of geographical distance and environment in beetle species composition at the macro scale. The pure effect of tree species and tree phylogenetic β-diversity was 10% and 6%, respectively, while the pure effect of the environment was 5% (Fig. 4-II) and 3% (Fig. 4-I) separately. However, the pure effect of geographic distance was very small at 1% (Fig. 4-II) and 0 (Fig. 4-I). The remaining unexplained variation of the main matrix was 28% (Fig. 4-II) and 31% (Fig. 4-I), respectively.
Through analysis of linear-mixed effect model, we found that the best model (i.e., delta AIC is equal to 0) retained both plant diversity and environment variables with significant correlation. All residuals of the models showed no significant spatial patterns (p = 0.8351), indicating that our mixed model explicitly incorporated the spatial dependence between plots, transects and regions .The best model showed that Fixed effects (i.e., lakes nested in regions) explained considerable variations of the models, with 65.08% (Tables 1). And random effect explained 32.28% (Tables 2). For the environment metrics, beetles standardized Shanoon diversity is positively correlated with AHR_stdz, AMH_stdz, MTCM_stdz and MTWM_stdz. But negatively correlated with ATR_stdz. And for the plant diversity metrics, beetles standardized Shanoon diversity is negatively correlated with PlaSimpson_stdz and PlaPD_stdz, but positively correlated with PlaChao1_stdz and PlaShannon_stdz.