Disentangling the determinants of and mapping the root-to-shoot ratio of Chinese forest ecosystems

The root-to-shoot ratio (R/S) is a key parameter for estimating forest carbon stocks and modeling carbon cycles at global and regional scales. We compiled a large dataset (n = 7,980) from Chinese forest ecosystems and employed the Random Forest algorithm, combined with shapley additive explanations (SHAP) and piecewise structural equation modeling, to identify the key drivers of R/S and their threshold effects. Consistent with optimal partitioning theory, R/S variation is driven by both biotic and abiotic resource limitations. Our results showed that the average R/S in natural forests (NF) (0.262 ± 0.001) was significantly higher than that in planted forests (PF) (0.228 ± 0.002), and deciduous forests exhibited higher R/S values than evergreen forests. Mean annual temperature (MAT) and stand age (Age) were the main drivers of R/S variation in NF and PF, respectively. The MAT threshold for R/S in NF was higher in broadleaved than in coniferous forests. In PF, R/S generally decreased with Age, with the transition zones from positive to negative responses occurring earlier in broadleaved forests (14.01–16.22 years) than in coniferous forests (17.31–19.08 years). Climate, soil, and stand factors exerted direct negative effects on R/S, whereas topography mainly influenced R/S indirectly through its effects on climate. The predicted R/S values ranged from 0.161 to 0.487, and the coefficient of variation attributable to forest type and forest origin was predominantly below 15%. These findings highlight the importance of forest origin and forest type in R/S estimation and provide a reliable reference for carbon accounting in China.