Allometric Models for Estimating Above- and Belowground Biomass for Coconut Palms in Tanzania

This study developed the locally calibrated allometric models for estimating above- and belowground biomass (AGB and BGB) of coconut palms ( Cocos nucifera ) in Tanzania as a function of diameter at breast height (D) and total height (H). Destructive sampling was conducted on 46 palms for AGB and 29 palms for BGB across two contrasting coastal districts: Mkuranga, located along the Indian Ocean, and Kisarawe, situated farther inland. Five nonlinear model forms were fitted for to both pools, and model performance was assessed using R², RMSE, AIC, and Mean Prediction Error (PE%). Leave-one-out cross-validation (LOOCV) was applied to evaluate predictive reliability given the limited sample size. Model results showed moderate explanatory power, with R² ranging from 0.62–0.75 for AGB and 0.52–0.53 for BGB. Contrary to conventional findings for dicotyledonous trees, H consistently outperformed D as the strongest predictor for both AGB and BGB. This pattern reflects the monocot growth form of coconut palms, where D stabilizes early while biomass accumulates primarily through vertical growth. Models combining H and D did not significantly improve predictive accuracy and often produced insignificant coefficients. LOOCV confirmed that H-only models yielded the lowest bias and most consistent predictive performance. The study provides the allometric equations for coconut palms in Tanzania, offering an essential tool for biomass quantification in coconut-based agroforestry systems and carbon accounting frameworks. These models represent a methodological advance over generic IPCC defaults and support a more accurate assessment of carbon stocks in coastal agricultural landscapes.