Effect of Local Site Quality Variation on Height-Diameter Allometry of Pinus patula at Sao Hill Plantation, Tanzania

This study developed height-diameter (H-D) models that explicitly incorporate site quality for Pinus patula in Sao Hill Plantation, in Tanzania, using a large dataset of 42,030 sample trees measured across 976 compartments and 21,355 plots. Site classes were assigned using dominant height-age curves from existing yield tables, resulting in four site-specific datasets used to fit nonlinear H-D equations. Models were estimated using the NLP Procedure in SAS, which calibrated both mean and variance structures to address heteroscedasticity. Model performance was evaluated using AIC, standard error, pseudo-R2, and mean prediction error (PE%). Results show that site-class-specific models consistently outperformed the general model, with higher R2 (0.63 - 0.74), lower SE (2.76 - 4.21), and non-significant PE% values. Asymptotic H parameters increased progressively from site class IV to I, confirming strong site-level influence on H growth. Validation using independent data demonstrated lower bias for site-specific models across diameter classes and site classes, whereas the general model introduced substantial errors, particularly in poorer sites. These findings highlight the limitations of a unified H-D model and demonstrate that integrating site quality significantly improves H estimations. The study provides site-class-specific H-D models for operational forest inventory, volume estimation, carbon accounting, and growth-and-yield planning in major pine-growing areas in Tanzania, particularly the Southern Highlands.