Gradient Boosting Models Predicting Standing Volume of Trees in an Urban Framework

Abstract

Reliable prediction of stem volume is crucial for effective management of urban trees along with their ecological assessment. Traditionally, regression models were used to estimate growing stocks of a forest, yet they often fall short when handling the complex, non-linear patterns typical of biological data, potentially introducing biases and errors. Tree stem volume, a critical metric in forest biometrics, is generally estimated through easily measured parameters such as diameter at breast height (DBH) and total tree height (H). This study investigates the utility of two machine learning (ML) techniques viz. Gradient Boosting Machine (GBM) and Extreme Gradient Boosting (Xgboost) to predict the stem volume of five moderate to fast- growing deciduous tree species of urban Delhi Forest viz., Cassia fistula, Bombax ceiba, Azadirachta indica, Albizia lebbeck and Tectona grandis, using basic field measurements, including additional tree-level predictors as well. Log- transformation of tree volume (target variable) was done to stabilise variance. Both models demonstrated a good fit (r ≥ 0.97, p < 0.001). Moreover, residuals displayed no signs of heteroscedasticity. The findings highlighted the potential of ML models to identify complex nonlinear patterns, thus improving the accuracy of forest inventory predictions, thereby supporting more effective and data-driven forest management strategies.