Abstract: Accurate allometric models are essential for quantifying forest carbon sinks. China’s extensive subtropical forests lack reliable biomass models, limiting carbon estimation. Here, we developed improved above- and below-ground biomass models using 1,993 harvest-based records from 166 sites. Various biomass models were developed, incorporating tree diameter at breast height (D, in cm), tree height (H, in m), and wood density (ρ, in g cm^-3). Model performance was rigorously evaluated against existing approaches using the coefficient of determination (R²), root mean square error (RMSE), mean percent bias (MPB), and mean absolute percent error (MAPE) at individual and plot scales. The optimal above-ground biomass (AGB, in kg) model, AGE=0.0526·(ρD²H)^0.9891 (R²=0.96, RMSE=62.18 kg, MPB=2.06% and MAPE=26.69%), and below- ground biomass (BGB, in kg) model, BGB=0.0107·(ρD²H)^0.9803 (R²=0.90, RMSE=13.96 kg, MPB=6.84% and MAPE=33.16%). Our AGB estimates showed superior accuracy to Chave et al.’s pan-tropical model, with lower systematic bias (MPB=-1.59% vs. 13.88%) and error (MAPE=27.34% vs. 31.36%), particularly for trees with D < 100 cm (~97% of total validation data). Root-to-shoot ratios differed between functional types (0.231 ± 0.084 for broadleaf versus 0.194 ± 0.062 for coniferous species) and shifted with spatial scale and soil type; the paired AGB–BGB equations improved belowground estimation relative to a single constant ratio. These models facilitate more robust pan- tropical carbon assessments, thus supporting informed forest management and global climate policy decisions, especially in regions previously understudied.

Key words: forest biomass models, pan-tropical trees, carbon, wood density, root-to-shoot ratio