Progressively synergizing individual-tree and area-based approaches for regional forest biomass mapping using aerial LiDAR

Aerial LiDAR provides essential three-dimensional structural information for accurate forest biomass estimation across large spatial extents. However, conventional approaches, including the area-based approach (ABA) and individual tree segmentation (ITS), face substantial cost constraints in model development and biomass mapping, respectively, limiting their operational deployment for large-scale forest inventories. We developed a progressively synergizing ITS–ABA strategy that reduces the costs required for biomass estimation model development by > 98% compared to traditional ABA while maintaining comparable accuracy. Rather than relying on extensive field plot inventories, we calibrated the ITS biomass models using a limited set of field-measured individual trees. Subsequently, the resulting biomass estimates derived from high-density aerial LiDAR data (>100 pt·m⁻²) via ITS were utilized as augmented training samples. The ITS–ABA method achieved strong agreement with field-measured values (R² > 0.81, rRMSE <20.00%) and matched conventional ABA model accuracy (|ΔR²| < 0.05, |ΔrRMSE| < 2.00%). For a 30,000 km² eucalyptus plantation monitoring case, our method required only a small number of high-density point cloud samples and limited single-tree measurements vs. 165 field plots for conventional ABA model development, achieving the 98% cost reduction. We validated this approach at both forest farm and provincial scales, where the ITS–ABA model outperformed the ABA model in cross-scale transfer from local calibration to regional deployment. This framework substantially reduces field requirements during biomass model calibration and supports accurate regional biomass estimation when combined with low-density aerial LiDAR for large-area deployment, demonstrating strong potential for operational plantation forest monitoring.