Comparative Evaluation of Nonlinear Regression and Neural Network Models for Fir Tree Form-Height Factor Estimation in Support of Sustainable Forest Management

The multifunctional role of forests includes estimating wood volume and biomass, which provide significant information for sustainable forest management. Nevertheless, because tree stems are irregular in shape, accurately estimating tree volume and biomass is difficult and requires determining the taper of the tree bole. This need has led to extensive research on factors that quantify tree stem irregularities. The goal of this study is to construct reliable models to estimate the stem taper of standing fir trees in the University forest at Pertouli, Trikala, Greece, using the form-height factor. Specifically, the generalized regression neural network and the resilient propagation artificial neural network were used for their effectiveness in identifying and learning complex patterns in the ground-truth data. These models were selected as alternatives to standard nonlinear regression to compare their accuracy, spanning from a traditional nonlinear regression method to more flexible artificial intelligence approaches. Overall, the resilient propagation artificial neural network model performed best, showing the lowest root mean square error (1.7237m), average absolute error (1.407m), and Furnival’s index of fit (1.7237m), indicating higher prediction accuracy and smaller errors. The generalized regression neural network model performed very similarly, with slightly higher root mean square error and average absolute error values, suggesting it is also a strong predictive model. In contrast, the nonlinear regression model had much larger root mean square error and average absolute error values, indicating weaker predictive accuracy. In conclusion, the machine learning methods tested can be recommended as a reliable alternative to support modeling and informed decision-making in forest management for sustainable forest ecosystems.

Cite this article as: Diamantopoulou, M. J. (2026). Comparative evaluation of nonlinear regression and neural network models for fir tree form-height factor estimation in support of sustainable forest management. Forestist, 76, 0132, doi: 10.5152/forestist.2026.25132.