Abstract
The distribution of assimilated carbon (C) among roots, stems, and leaves is a central process in terrestrial ecosystem dynamics. Yet the biomass allocation schemes used in current global vegetation and land surface models pre‐date the existence of large plant‐trait data sets and remain largely untested. Here we formulate hypotheses on the controls of root: shoot biomass ratios (R:S), based on eco‐evolutionary optimality principles, and assess them quantitatively by analysing data on nearly 30,000 observations of R:S. We analysed global R:S patterns using multiple linear regression models for woody and herbaceous species separately, considering as candidate predictors growing‐season mean temperature ( T g ), gross primary production (GPP), a measure of root‐zone water capacity (RZ WC ), soil pH, sand content, aridity index (AI), and plant traits: vegetation height (H), leaf thickness (LT), leaf dry matter content (LDMC), specific leaf area (SLA), specific root length (SRL), and rooting depth (RRD). R:S was systematically greater in herbaceous plants. R:S decreased with T g , GPP, and height but increased with sand content, RRD, and LDMC in both woody and herbaceous plants. However, AI and leaf thickness had opposing effects on R:S. RZ WC and SLA were important in woody plants, while pH and SRL played a larger role in herbaceous plants. The models explained 13% (woody) and 31% (herbaceous) of R:S variation. The lower explanatory power for woody plants is likely influenced by unmeasured variations in (for example) forest age and canopy position. These empirical findings provide a step towards a quantitative theory of plant C allocation responses to resource availability and an improved C allocation scheme for ecosystem models.