Structural equation modeling
Given the high collinearity among the climatic and edaphic factors
(Figure S4), we extracted the first PCA axis for both sets of variables:
PC1 explained 94.36% and 68.99% of the variance for the climatic and
edaphic variables, respectively (Table S4, S5). The final piecewise SEM
(standardized path coefficients provided in Table S6; Figure S5), which
adequately fit the data (Fisher’s C = 10.302, d.f. = 8, P= 0.244; AICc = 54.302), explained 63.7% of the
variance in species’ turnover (R2 = 0.637) and 23.1%
of the variance in intraspecific variability in herbivory
(R2 = 0.231) (Figure 3). In the final piecewise SEM,
intraspecific variability decreased with the climatic PC1 (standardized
path coefficient β = -0.265, P = 0.016), while it was not
affected by the plant community composition DC1 (β = 0.128, P =
0.222) or soil PC1 (β = 0.023, P = 0.808); this suggests that
climatic effects on intraspecific herbivory variability drove changes in
community-wide herbivory along the study latitudinal gradient. The plant
community composition DC1 (β = -0.251, P = 0.222), but not the
climatic PC1 (β = -0.091, P = 0.487) or soil PC1 (β = 0.098,P = 0.291), decreased species’ turnover effects.