Fig. 2: Soil hydraulic conductivity controls critical soil water thresholds of ecosystem water limitation.

a, Global critical soil moisture thresholds correlated with the mean (range indicated by horizontal error bars) gravimetric sand fraction per soil textural class (linear regression of θ_crit to mean sand fraction for all observations—that is, FN + SFN—for soil texture specific maximum and minimum θ_crit, and for all FN and SFN θ_crit separately, are shown as solid black, solid grey, dashed–dotted grey and dotted grey lines, respectively). b, Critical soil moisture thresholds are well predicted by applying the model to the site-specific soil textural information (linear regressions of observed-to-simulated θ_crit, considering all soil textures, and by excluding clay soils, are shown as solid black, while linear regressions of FN and SFN separately are shown as dashed–dotted and dotted grey lines, respectively). c, The sand fraction dictates the span of observed (FN, SFN) and simulated (sim.) critical soil moisture thresholds (S_θcrit) and the difference between maximum and minimum θ_crit per soil textural class. S_θcrit is a measure of the θ_crit variability of the observations (black) and of the θ_crit sensitivity to the model parameters, such as varying soil hydraulic properties in a soil textural class (brown), and varying plant traits (plant vulnerability, ψ_x50, and root length, L_root, in green). The negative slopes of the linear regressions of S_θcrit (except for L_root) demonstrate the decreasing variability and sensitivity of θ_crit with sand fraction, in both observation and simulation (solid black and coloured dotted lines, respectively). d, The observed median critical soil water potential thresholds (ψ_crit) for FN (open circles) and SFN (crosses) confirm the expected decline with increasing sand fraction (note the simulations (filled circles and red solid curve) correspond to Fig. 1d, and the log₁₀-transformed linear regressions for all ψ_crit, not only the medians, are shown as solid black, with FN and SFN shown separately as dashed–dotted and dotted grey lines, respectively). a,b, The data are presented as grouped (FN, SFN) box plots (the thick solid line represents the median, the lower and upper hinges correspond to the first and third quartiles, with the whiskers extending to the highest or lowest value, respectively, but no further than 1.5 times the interquartile range, and the width of the boxes scales with the square root of the number of observations in each soil textural class) in combination with individual observations displayed as points along the boxes (n_FN and n_SFN  indicate the number of FN and SFN θ_crit observations per soil textural class, respectively, while the number of sites per soil textural class is given in Supplementary Table 2). a–d, Adjusted R² values (R²_adj) and two-sided P values of the linear regression slopes (regression t-test) are indicated. *P < 0.05, **P < 0.01, ***P < 0.001. NS, not significant (P > 0.05).