Fig. 2: Root bimodality predicted by site-level factors.

a Feature importance ranking from a random forest classification model (Methods) identify total fine root biomass (g/m²) as the most important factor in predicting the modality of a site. Of these 12 factors (see Methods, site-level factors), land cover is the only categorical variable, the rest are continuous. The features in the orange bars are highlighted in (b, c). b Unimodal sites (n = 35, 990 g/m²) have nearly 3 times more root biomass than that of bimodal sites (n = 9, 368 g/m2; two-sided Student’s t-test, p = 0.004). Note that y axis is on logarithmic scale. c Site modality (bimodal vs. unimodal) are not independently distributed across vegetation types (forest in green, grassland in blue, shrubland in purple, cultivated in red). Bimodal sites are more frequently associated with shrubland but less with grassland (two-sided chi-square test, p = 0.019, χ² = 9.6). In each box plot, the lower and upper bounds of the whiskers denote minima and maxima, the center line denotes the median, and the lower and upper bounds of the boxes represent the 25% and 75% quantiles, respectively. Source data are provided as a Source Data file.