RESULTS
Our defaunation model included data from 58 species from 53 genera in 30 families (Table S2). With 10,000 iterations per species, this created 58,000 ‘defaunation response’ values across all defaunation scenarios.
In our additive model, the most important variable for explaining defaunation response was seed predation (model coefficient β = -0.153), explaining 6.82% of model variance, followed more distantly by seedling trampling (β = -0.076, 1.69% of variance; Figure 2). Across all species, changes in seed predation predicted defaunation response (Figure 3A). No phenotypic traits were associated with defaunation response across species.
Important statistical interactions emerged between seed predation and (i) angiosperm versus gymnosperm (β = 0.146), (ii) seed mass (β = 0.097), and (iii) specific leaf area (β = -0.073; Figure S2). When defaunation increased seed predation, populations of angiosperms (Figure 3B), small-seeded species (Figure 3C), and taxa with very high specific leaf area (Figure S3) declined. Gymnosperms, large-seeded species, and taxa with low to medium specific leaf area were relatively unaffected by defaunation regardless of seed predation.
Several phenotypic traits were significantly correlated with species vital rates (Table S6). Seed mass, in particular, was positively related to juvenile survival (r = 0.419, P = 0.001) and adult survival (r = 0.333, P = 0.011), but was negatively related to adult fecundity (r = -0.354, P = 0.006). Fecundity also correlated with whether a species was an angiosperm or a gymnosperm (r = 0.480, P< 0.001) and was positively related to tree height (r = 0.339, P = 0.009).