HARNESSING ARTIFICIAL NEURAL NETWORKS FOR ENHENCED WOOD VOLUME ESTIMATION FOR EUCALYPTUS CAMALDULENSISTEWNDS IN USMAN DANFODIYO UNIVERSITY ,SOKOTO

Abstract

on (R²). The ANN model outperformed traditional models, achieving RMSE values of 0.0201 (training), 0.0150 (validation), and 0.0216 (testing), with the highest R² (0.7290) during testing. These results demonstrate the superior accuracy of ANNs in modeling nonlinear relationships and highlight their potential to advance precision forestry and sustainable resource management. Future studies could explore integrating remote sensing data for broader application.

Accurate wood volume estimation is essential for sustainable forestry management, supporting resource planning, carbon sequestration analysis, and economic evaluation. Traditional regression models often struggle to capture the complex, nonlinear relationships between tree metrics and wood volume. This study evaluates the efficacy of Artificial Neural Networks (ANNs) for wood volume estimation in Eucalyptus camaldulensis plantations at Usmanu Danfodiyo University, Sokoto, with three specific objectives; which includes; establishing the relationship between the measured variables; Training the Artificial Neural Networks (ANNs) Model to estimate wood volume, and comparing its performance with traditional regression models. Using a complete enumeration approach, data on Diameter at Breast Height (DBH) and Total Height (TH) were collected from 142 trees. Regression analysis, including both simple and multiple linear models, were employed to establish the relationship between the measured variables. DBH was identified as the primary predictor of wood volume, explaining 69% of its variance (R² = 0.6892), while TH accounted for 7.66%. A combined regression model improved prediction accuracy (R² = 0.7365). Regression models including Chapman-Richards, Gompertz, Exponential, and Logarithmic models were developed alongside an ANN model implemented in MATLAB 2024b. Model performance was assessed using Root Mean Square Error (RMSE) and Coefficient of Determinati