Accurate Subtropical Mixed Forests Volume Estimation Through UAV-LiDAR and Random Forest

This study refines the application of unmanned aerial vehicle light detection and ranging (UAV-LiDAR) for forest parameter extraction, focusing on calibration aspects of random forest models within subtropical mixed forests. A notable inverse correlation was identified between high-density LiDAR-derived intensity percentiles and stand volume of 20 m × 20 m plots. Using a random forest model with a 20% validation proportion (30 plots), we achieved a validation coefficient of determination (R2) of 0.61 and root mean square error (RMSE) of 27.2%. Key variables included radar echo intensity and forest gap fraction, emphasizing their importance in volume estimation. Additionally, analyses highlighted forest gap fraction as an important variable in model construction, suggesting careful consideration of its distribution in sample plot selection.