Machine learning-based stem taper model: a case study with Brutian pine

Abstract

Stem taper models are essential tools in forestry, allowing for the estimation of stem diameter at any height, as well as the calculation of merchantable and total stem volumes and wood assortments along the tree bole. Therefore, accurate taper prediction is crucial for sustainable forest resource assessment. This study developed stem taper models for estimating tree diameter using both traditional regression and machine learning (ML) approaches, using Pinus brutia Ten. as a model species. The research focused on two machine learning techniques, Random Forest (RF) and Extreme Gradient Boosting (XGBoost) to predict stem taper in comparison to traditional taper models. A total of 121 destructively sampled trees were measured for stem diameter at multiple heights, and various taper models were evaluated for their accuracy. The results show that the XGBoost model outperforms all other approaches, demonstrating superior predictive accuracy with minimal error, as indicated by lower root mean square error (RMSE), mean absolute error (MAE), and bias values. While RF also performed well, XGBoost was selected for this study due to its better predictive performance and the more consistent error distributions between the training and test datasets. This research highlights the potential of ML techniques in forest modeling, offering enhanced accuracy and efficiency for forest inventory and management applications.