Machine learning study demonstrates accurate multi-species tree age estimation across stem discs, indicating radial growth rate reduces reliance on explicit species data.
Key Points
To develop a generalized multi-species tree age estimation framework using trunk radius and approximate relative radial growth rate within machine learning and ensemble models.
Analyzed 186 stem discs from four tree species collected at a sampling height of 1.3 m to extract trunk radius, age, average annual-ring width, and relative radial growth rate (RGR).
Trained four machine learning models—Backpropagation Neural Network, Random Forest, Extreme Gradient Boosting, and Support Vector Regression—and combined them using a weighted ensemble strategy.
The weighted ensemble model achieved an R² of 0.865, a prediction accuracy of 85.5%, an RMSE of 4.318 years, and an MAE of 3.058 years.
Incorporating RGR improved performance over a model using trunk radius and tree species alone, raising R² from 0.589 to 0.865 and reducing RMSE from 7.345 to 4.318 years.