Fig. 4: Our model correctly predicts the response of χ to vapour pressure deficit.

a, The model-predicted distribution of the slope of the relationship between logit(χ) and log(D) for the analysed species (grey bars) is well within the range reported in ref. ⁴⁶ (their reported mean and confidence interval is shown by the green line and green region, respectively). It is significantly different from −0.5 (orange line; a one-sample t-test shows a predicted mean slope of −0.7 and a 95% confidence interval of (−0.72, −0.68)). For each species, we calculate the predicted slope by varying vapour pressure deficit in the range 5–5,000 Pa while keeping other environmental parameters constant (at values reported in the respective experiments, with ψ_s = 0) and using fitted trait values (Table 1). b, This slope is correlated with the ψ₅₀ (black points are species-specific values and the blue line is a linear regression line), with more negative slopes observed for species with less negative ψ₅₀ (drought avoiders). This could be a reason why earlier datasets supported a slope value of −0.5, as such datasets were often dominated by temperate evergreen species, which are typically characterized by highly negative values of ψ₅₀.