ABSTRACT:

Accurate modeling of the height-diameter (H-D) relation is significant for effective forest management, especially in tropical plantations where data on height are normally limited. In this study, the relative efficiency of ten three-parameter non-linear models to predict tree height from diameter at breast height (DBH) was evaluated for Gmelina arborea plantation in Southwestern Nigeria. The models were also compared and contrasted on the basis of various statistical measures, such as; Root Mean Square Error (RMSE), Coefficient of Determination (R²), Mean Absolute Error (MAE), Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), residuals plot and Dense Ranking to determine their goodness-of-fit and predictive accuracy. Results were manifested with slight variation in performance of the models, of which the Korf H-D model was most suited for height prediction due to its overall better accuracy across all measures of 2 evaluation. It had the smallest RMSE, BIC, AIC, MAE and the highest R of 3.918, 1630.585, 1617.445, 3.046 and 0.816 (81.6%) respectively. It had high compatibility of model complexity with predictive ability, and thus it was a potential candidate for use in real application in forest inventory and growth prediction. Although the other models also gave a good explanation of the dataset, the Schnute H-D model was the poorest, manifested with the poorest fit and highest error rates among all the models tried. The study highlights the significance of correct model choice in forest modeling and recommends the employment of the Korf H-D model in estimating tree height in Gmelina arborea plantation in similar ecological regions. Future research can explore the integration of additional site variables or the application of mixed-effects and machine learning models to enhance predictability