Height-Diameter Allometry of Trees in Sacred Grove Forest in North Central Nigeria

Abstract

This study examined the diameter at breast height (DBH), total tree height (THT) of trees and modelled the height-diameter across nonlinear regression (NLR) and artificial intelligence (AI) approaches in the sacred grove forest, Benue State, Nigeria. The study reported a mean DBH of 40.92 cm and THT of 18.76 m, indicating a structurally diverse and uneven-aged forest. The DBH distribution was positively skewed (1.08) with moderate kurtosis (0.41), while tree height was nearly symmetric (skewness = 0.08) and slightly platykurtic (–0.34). DBH and THT correlated with each other in a curvilinear manner, where the height increment was decreasing after a point of about 50 cm DBH. Out of the seven fitted NLR models, the Weibull model (M2) best overall statistical model with a coefficient of determination (R 2 = 0.89), a minimum Akaike Information Criterion (AIC = 453.3), and an acceptable residual distribution, followed by the Chapman Richards model (R 2 = 0.857, AIC = 453.9). The Michaelis-Menten model was also competitive and exhibited low residual bias and biological realism. The XGBoost algorithm was the most predictive and least biased in the AI models, with R 2 = 0.865, RMSE = 1.623 m, MAE = 1.210 m, and MAPE = 9.37%. Thus, this study shows that AI and the classic regression methods can efficiently estimate the height of trees using their diameter; however, the XGBoost model AI-based gave a better performance in the conditions experienced in the heterogeneous forest.