A compatible model for aboveground biomass of individual culms in

Introduction: Methods: We addressed this knowledge gap using a destructive sample of 45 culms (2- and 3-year old culms) harvested from a managed plantation in Sanya, Hainan, China. Two modeling strategies were compared: (i) independent models fitted by weighted non-linear least squares and (ii) compatible models based on Seemingly Unrelated Regression (SUR) that guarantee additivity among culm, branch, and leaf biomass. Predictor variables were diameter at breast height (DBH), total height (H), and a combined variable (DBH)²H. Model performance was evaluated using 1,000 bootstrap iterations with out-of-bag (OOB) predictions, and final model parameters were estimated from the full dataset. Results and discussion: 2-year-old culms were roughly twice as large in DBH and H, and contained fourfold more AGB than 3-year-old culms. Independent models demonstrated that DBH was the most effective singular predictor for each component and AGB. Compatible SUR models achieved similar accuracy (R² > 0.9 for AGB and culms, > 0.7 for leaves, > 0.5 for branches) while reducing the discrepancy between summed components and total AGB to virtually zero. We developed the first SUR-based compatible biomass models for