Developing an optimal individual tree diameter growth model for uneven-aged Pinus yunnanensis forests using machine learning algorithms

Abstract

The objective of this study was to develop more accurate predictions of the diameter growth of Pinus yunnanensis and to analyze the impact of various factors on its diameter growth, providing valuable management recommendations for forest management. To this end, various machine learning methods were employed to construct individual tree diameter growth models for P. yunnanensis. The research was based on single-period survey data and core sample data from 11 permanent plots in Cangshan mountain, Dali, Yunnan Province. In addition, the impacts of tree size, competition, site quality, and climatic factors on the growth of P. yunnanensis diameters were considered. Four machine learning methods were employed to develop the models: Random Forest, XGBoost, Multilayer Perceptron, and Stacked Multilayer Perceptron (Stacked-MLP). The models were evaluated and compared using a k-fold strategy, based on the coefficient of determination, Root Mean Square Error, and Mean Absolute Error. The results of the fivefold cross-validation demonstrated that the Stacked-MLP model exhibited the highest performance, with an R2 of 0.8508, RMSE of 0.2907 cm², and MAE of 0.1928 cm². The feature importance methods from Random Forest, XGBoost, and SHAP analysis indicated that competition and tree size were the primary drivers of tree growth, while climate and site factors had a more limited impact in explaining variations in tree growth on a small, local scale.