Fig. 3: Tests of correlation mechanism hypotheses described in Fig. 1.

All richness values are from the consensus richness maps produced to combine the richness predictions published by different authors (see ‘Methods’). Evidence for symbiosis-driven diversity coupling effects: A the partial percentage deviance in fungal richness directly explained by plant richness at the global level (based on generalised additive models (GAMs) including environmental covariates using 10,000 random global grid cells, see Supporting Information Fig. S6), and B partial residual plots from GAMs testing how richness correlations within ecoregions (points) changes as the percentage of vegetation biomass belonging to potential host plant species³³ increases (AM host vegetation for AM fungal–plant correlations and ECM vegetation for ECM fungal–plant correlations). Evidence for environmental covariates driving diversity correlation patterns: C p-values show results of two-sided chi-squared tests testing the hypothesis that positive plant–fungal correlations (blue) occur in ecoregions when plant and fungal richness respond similarly (top right and bottom left panel segments) to mean annual temperature, mean annual precipitation, and soil pH, and negative plant–fungal correlations (red) occur when they respond in opposite directions (top left and bottom right panel segments). Coefficient estimates were calculated from linear models including plant richness, fungal richness, and percentage host plant vegetation biomass as covariates to remove any variation due to potential symbiotic effects (see ‘Methods’). Biome summaries are provided in Supporting Information (Fig. S4). Evidence for both long-term and short-term legacy effects influencing correlation patterns: partial residual plots for GAMs testing the relationship between richness correlations within ecoregions and D climate stability index²⁴ and E human development percentage (see ‘Methods’). Variables shown in (B, D, E) were included in the same GAMs, along with other environmental covariates relating to climate, soil properties, vegetation, and species richness (see ‘Methods’ and Supporting Information Fig. S5). The line shows the fitted curves, and shading shows one standard error. AM arbuscular mycorrhizal fungi, ECM ectomycorrhizal fungi.