Changes of the relative importance of diversity vs. stand factors with grain size
The models in Table 3 were based on (sub)plots that differed markedly in sample size, consequently we randomly extracted a same number of subplots for a better comparison across grain sizes (see Methods). Figure 1 again showed that, for forest biomass and productivity, stand factors revealed a markedly stronger effect than diversity indices at each grain size (especially for biomass), no matter measured by the pure or total effect. This confirms our conjecture that stand factors are dominant in driving forest biomass and productivity. However, we also noted that the total and pure effects of diversity on biomass both showed an increase from the 400 to 1200 m2 grain size (for the 2500 m2 results, see below). For productivity, the total effect of diversity increased with grain size while the pure effect decreased, suggesting that the joint effect between diversity and stand factors increased with lager plot size. We thus concluded that the biodiversity effect on forest biomass and productivity might increase with grain size.
For the stability of biomass and productivity, however, the total and pure effects of diversity were markedly higher than that of stand factors across grain sizes, confirming that diversity is dominant in affecting ecosystem stability. However, the total and pure effects of diversity and stand factors on the two stability metrics did not revealed a consistent pattern across grain sizes.
For comparison with the above results based on random sampling of subplots, we also reported the variation partitioning analyses based on models in Table 3 (Fig. S3). We found that while the basic results were similar (stand factors were important for biomass and productivity, while diversity was dominant for stability), there were also notable differences. For biomass and productivity, diversity occasionally showed higher explanatory power in Fig. S3 compared with that in Fig. 1, while at other grain sizes diversity was excluded from the models in Fig. S3. Meanwhile, the explanatory power of diversity from 400 to 1200 m2 in Fig. S3 were ca. 23% and 20% for biomass and productivity, respectively, which were clearly lower than that in Fig. 1 (ca. 40% and 37%, respectively). These differences suggest that the results based on random sampling of plots may be more robust (see Discussions for details). Consequently, the results of the 2500 m2 (which was not based on random sampling) should be viewed with caution, and was listed in Fig. 1 for reference only, and not reported in subsequent analyses.
The relative importance of different diversity components and dimensions
The results based on random sampling of the subplots showed that (Figure 2), functional diversity (including the effect of richness, evenness and divergence) seemed to be more important for biomass and productivity than phylogenetic and taxonomic diversity, and the importance of functional diversity increases markedly with larger grain size. Thus the increase in the explanatory power of diversity in Figures 1A and 1B may be mainly caused by functional diversity. For biomass stability, functional diversity is also clearly more important than phylogenetic and taxonomic diversity, while the three dimensions of diversity showed similar importance on productivity stability. Meanwhile, the relative importance of three diversity dimensions did not reveal a consistent trend across grain sizes, which may be why the effect of diversity did not show a clear pattern with plot size in Figures 1C and 1D.
As for the three components of diversity (Figure 3), the importance of richness (mean importance of species, functional and phylogenetic richness) on ecosystem functions and stability were generally higher than that of evenness and divergence. However, for biomass and productivity, the importance of divergence increase with grain size. While for the two stability measures, the change of importance value with grain size differed among three diversity components.