Extended Data Fig. 8: Partial residual plots of the indirect species richness- and species turnover-mediated effects of environmental heterogeneity on the contribution of diversity to variation in ecosystem functioning.

a, Topology of the Bayesian hierarchical structural equation model with the strongest support. Solid paths were supported by the Bayesian variable selection, whereas dotted paths were not (see Extended Data Fig. 7). Plots b–e visualize partial relationships indicated by the bold letters of the structural equation model in (a). Relationships of (b) variation in species richness and (c) species turnover with environmental heterogeneity. Relationships of the diversity effect with (d) variation in species richness and (e) species turnover. All variables were scaled to zero mean and unit variance before analysis. Units on the y-axes are standardized residual deviations from predicted partial scores after conditioning on all predictor variables except for the one shown on the x-axis and after conditioning on the random effects (function ID, n_function = 22). The colors of the circles represent the two types of comparisons (orange, within ecosystem types; black, across ecosystem types). The grey arrows indicate the directional change in the predictor and response variable with increasing environmental heterogeneity (that is, from comparisons within to comparisons across ecosystem types). The black lines depict the partial regression slopes. Sample size is n = 43 (n = 21 within and n = 22 across ecosystem types, respectively). Note that the within-ecosystem comparison for biomass stocks of microorganisms is missing, because data were only available for one replicate per ecosystem type.