Statistical estimation the effect of edaphic, topography, DBH, genetic variation and spatial on MVPS
The relative influence of microhabitat, individual life stage, genetic variation and spatial attributes on MVPS were determined by using MRM in R 3.1.0 with the “ecodist” packages (Goslee & Urban, 2007). MRM, an extension of partial Mantel analysis, can employ non-parametric or non-linear multiple regression. In this study, independent variables increasing one by one from single edaphic variables to multiple variables, such as edaphic, topography, spatial, individual life stage and genetic variation were used to construct five MRM models. We first used Spearman correlations to assess the relationships between environmental distance matrices and the MVPS matrix; then the significant explanatory variables were identified using a forward-selection procedure; finally, the significance of MRM models was tested with 1,000 permutations on the objects of the response distance matrix. The r2 from MRM analyses was used to assess MVPS variation explained by these significant environmental variables, and the significantly correlated variables were excluded within each group of variables (edaphic, topography, individual life stage, genetic variation and spatial variables). The relationship between DBH and methylation level was also analysed with MRM.