Experimental evidence supports the notion that more diverse communities have higher ecosystem productivity. Yet many gaps remain in our characterization of this biodiversity-ecosystem productivity (BEP) relationship with respect to non-grassland systems and dimensions of biodiversity beyond species richness (SR). We report here on early BEP findings from a 12 species forest diversity experiment established in 2013 at the Cedar Creek Long Term Ecological Research site in central Minnesota. We measured height and basal diameter of all experimental trees (N=8,960) from 2013-2015 and allometrically estimated annual incremental change in aboveground stem biomass. We first present a statistical model of overyielding in productivity using only biculture plots to disentangle the effects of species richness from those of phylogenetic (PD) and functional diversity (FD) using Loreau and Hector’s (2001) method to partition overyielding into complementary (non-additive) and selection (additive) effects of diversity on productivity. We subsequently present analyses from the full 140-plot experiment (including 5- and 12-species plots). We also assessed the scale-dependence of our findings by modeling BEP patterns at two scales: 1) at the plot level (64 trees per plot; 9.25 m2) and 2) at the neighborhood level using 8 individuals surrounding focal trees (1 m2).
Productivity was not predicted by PD and FD when we modeled BEP in bicultures only; instead, relative proportions of constituent species best predicted productivity. Models of productivity across all polycultures included SR and community weighted means of key traits (CWMs) rather than species proportions, and also did not include PD or FD. When all polycultures were considered, plots that were either more speciose or contained less shade tolerant trees with thinner, less nitrogen-rich leaves showed the highest overyielding in productivity. This finding suggests that plots dominated by multiple early-successional species were not only more productive, but also became more productive at higher diversity than did plots dominated by later-successional species. Generally, species richness, relative abundances, and functional identities, but not PD or FD, best predicted overyielding. Across all plots, overyielding was a consequence of complementarity, arising from species interactions rather than the effects of individual species. Selection effects were weak and idiosyncratic. Finally, the positive relationship between SR and overyielding in stem biomass in the plots disappeared when we considered neighborhoods. At the neighborhood scale, proportions of constituent species best predicted overyielding, suggesting that BEP patterns may be more pronounced at the plot rather than the neighborhood scale.