<p>Descriptive statistics for sampled trees.</p>

Accurate estimation of tree biomass and volume is essential for carbon accounting and climate reporting. This study developed site-specific allometric models for predicting stem volume and aboveground biomass of Gmelina arborea in the high-elevation Kwahu Highlands of Ghana. The trees were grouped into three DBH classes. DBH, total height, merchantable height, bole height, and crown dimensions of 45 trees were measured. Twenty-four of these were randomly destructively sampled for volume and biomass estimation. A range of linear, power, and log-linear regression models were evaluated to identify the most accurate predictors. Tree structural attributes changed systematically with DBH. As DBH increased, total height increases rapidly, bole height increased more gradually, and crown length showed strong growth. Expanded log-linear models [ln(Y) = a + b₁*ln(x) + b₂*ln(h) + b₃*ln(p)] incorporating DBH, total height, and wood density provided the best fit for both stem volume and biomass (R² > 0.90), with minimal error and strong agreement with observed values. Simpler DBH + height models showed significant deviations for biomass estimation but performed adequately for volume prediction. These findings highlight the importance of integrating multiple predictors in biomass models for unmanaged highland plantations. The resulting equations improve carbon stock estimation accuracy, supporting climate policy and sustainable forest management in tropical highland ecosystems.