Modeling diameter at breast height of Chinese fir (Cunninghamia lanceolata) using UAV LiDAR data in Southern China

Large-scale prediction of tree diameter at breast height (DBH) using airborne LiDAR remains constrained by models that inadequately address differences in tree growth stages and regional ecological variation. Existing approaches often overlook non-linear growth patterns and hierarchical spatial effects, thereby limiting predictive accuracy and scalability. In this study, we developed a DBH estimation model tailored for <i>Cunninghamia lanceolata</i> forests by integrating field-measured DBH with corresponding airborne LiDAR data collected from 26,768 trees across 130 plots in Guangdong Province, China. To capture growth-stage variability, a dummy variable approach was implemented to enable stage-specific adjustments within the model. Moreover, a two-level linear mixed-effects model was employed to account for nested spatial heterogeneity at both regional and stand levels. Competing model structures were rigorously evaluated using Akaike Information Criterion (AIC) and multiple error metrics, and the final model performance was validated with an independent dataset. Our results demonstrate that incorporating growth-stage differentiation and multilevel random effects significantly enhances model accuracy, with additional improvements observed upon including stand density and crown width indicators. The final model outperformed traditional approaches, effectively capturing spatial and ontogenetic variability. This study provides a methodological foundation for improving DBH estimation of <i>Cunninghamia lanceolata</i> using airborne LiDAR data. While further validation is needed, the modeling framework may also offer a potential basis for future applications using UAV-borne LiDAR platforms in similar forest environments.