Improving Regional Forest Biomass Estimation With a UAV-Driven Hierarchical Framework: A Case Study in Chinese Fir Plantations

Accurate estimation of forest aboveground biomass (AGB) is vital for carbon accounting and sustainable forest management, yet it remains constrained by the scale mismatch between field plots and satellite imagery. To address this challenge, we developed and validated a UAV-driven hierarchical framework integrating a Stacking Ensemble Learning (SEL) strategy for high-precision regional AGB estimation in structurally complex Chinese fir (Cunninghamia lanceolata (Lamb.) Hook) plantations. First, by exploiting the synergistic interactions among height, intensity, density, and vegetation metrics, the plot-scale SEL model achieved high precision estimates (R2 = 0.88, RMSE = 25.50 t/ha). These fine scale AGB maps were then aggregated using a polygon-to polygon upscaling strategy, generating robust training samples at the satellite pixel scale. Subsequently, these samples were used to train a regional SEL model that fused multi-source datasets— including Sentinel-1 SAR textures, Sentinel-2 spectral indices, and SRTM-derived terrain variables—to produce a regional AGB map. The proposed Ground-UAV-Satellite hierarchical framework achieved superior regional-scale accuracy (R2 = 0.65, RMSE = 35.83 t/ha), significantly outperforming the conventional Ground-Satellite method (R2 = 0.46) by increasing the explained variance by approximately 40% and reducing the RMSE by 21%. This research establishes a scalable and transferable methodology that mitigates sampling bias and scale-related uncertainties, offering a robust tool for large-area, operational forest-carbon monitoring.