Time and density simulations

For the second phase of the simulation process (see Fig. 3E), we used the estimated posterior range of each Trec parameter (see Table ​Table1)1) to estimate the time needed for the stabilization of the spatial genetic structure (SGS) of T. recurvata populations (hereafter referred to as 'time simulation’) and the tree densities values for which the lowest and highest SGS values are achieved (hereafter referred to as ‘density simulation'). To reduce the bias effect inherent to dynamic landscapes (e.g. with growing and falling trees), we performed both time and density simulations in static landscapes. We performed 2500-time simulations that emulate 50 years of spreading of T. recurvata individuals on the empirical landscape representation (Fig. 1), and 5000 density simulations of T. recurvata populations in ca. 0.40 hectares landscapes with five to 150 trees (maximum amount of trees that occupy the landscape) with random heights and same crown sizes. Yearly (for the time simulations) and after 30 years (for the density simulations), we recorded the individual and MLL abundances, as well as the MLG, MLL, and ramets’ origins (from clonal growth or sexual reproduction; hereafter referred to as ‘clones’ and ‘seeds’ individuals) of up to 15 random T. recurvata individuals established on each host tree representation of the empirical study (for the time simulations) or on up to 15 random trees for density simulations (for the density simulations). Based on the MLGs of each simulation, we calculated the average number of alleles (A), the expected heterozygosity (H_E), the inbreeding coefficient (F_IS), and the global subpopulation differentiation (F_ST), using the arlsumstat console version of Arlequin 3.1; the MLG and MLL turnovers among subpopulations (T/O_MLG and T/O_MLL), using the R package vegetarian^104,105; and the S_p statistics considering seven distance classes as in the empirical study⁶⁶ using the R package RClone¹⁰².