Crown-based biomass indicators and dual topographic regulation of carbon allocation in alpine Rhododendron shrublands on the Tibetan Plateau

Alpine Rhododendron shrublands are a key component of the carbon pool in the cold, high-altitude ecosystems of the Qinghai-Tibet Plateau; accurate estimates of their biomass and an understanding of the mechanisms underlying its distribution are crucial for comprehending vegetation adaptation and carbon cycling processes. Based on data from 63 plots across 21 sampling sites in the Sygera Mountain region (elevation 4,300–4,600 m), this study developed a Random Forest (RF) model for individual tree biomass calibrated using Gaussian Process Regression (GPR). The study also employed power-law regression, cluster analysis, and structural equation modeling to analyze biomass allocation strategies and their environmental drivers. The results show that the initial Random Forest model has an R² of 0.75, and the residuals exhibit weak but significant spatial autocorrelation; After GPR calibration, the model’s generalization performance improved significantly (CV R² = 0.834, RMSE = 61.66 g). Crown width and volume are the most important predictors, while the direct contribution of environmental factors is relatively weak. There is a significant allometric relationship between aboveground and belowground biomass, with a power-law slope of 0.633, indicating that aboveground growth is relatively prioritized. The root-to-shoot ratio data show that the warm, humid sampling sites on the southern slope are predominantly characterized by low R/S values, while the cold, humid sampling sites on the northern slope are predominantly characterized by high R/S values. Structural equation modeling further indicates that slope aspect and its interactions with elevation and wind speed significantly regulate vegetation structure; aboveground biomass is primarily mediated by vegetation structure, whereas belowground biomass is influenced not only by vegetation structure but also by the direct effects of slope aspect. This study provides a generalizable modeling framework for the spatial prediction of alpine shrub biomass and reveals the dual regulatory mechanisms of microtopography on carbon allocation in alpine shrublands, These findings may inform monitoring strategies and biomass assessment in alpine shrublands.