UAV LiDAR-Assisted Multi-Source Remote Sensing Estimation and Spatiotemporal Analysis of Forest Carbon Storage in Fuzhou City
Abstract
Accurate estimation of regional forest carbon storage is essential for assessing forest carbon sink capacity and supporting climate change mitigation. This study developed a UAV-LiDAR-assisted multi-source remote sensing approach to estimate forest AGB and carbon storage in Fuzhou City. UAV-LiDAR data acquired close to satellite overpasses were used to generate temporally matched plot-scale AGB samples through individual-tree segmentation, tree height–DBH estimation, and allometric equations. After screening, 204 samples were retained. Four predictor combinations (Landsat, SAR, Landsat + SAR, and Landsat + SAR + other factors) were combined with SWR and CatBoost to build eight AGB models after RF-RFE feature selection. The optimal model was applied to analyze forest carbon storage changes from 2015 to 2023. CatBoost generally outperformed SWR, with lower RMSE, MAE, and rRMSE. The CatBoost model integrating Landsat, SAR, and other factors performed best (RMSE = 19.83 t·hm−2, MAE = 17.02 t·hm−2, rRMSE = 21.73%). Forest carbon storage increased from 32.43 × 106 t in 2015 to 36.76 × 106 t in 2023, alongside increases in forest area and carbon density. These findings suggest the potential of UAV-LiDAR for temporally matched AGB sampling and regional forest carbon storage estimation in subtropical areas, contributing to relevant SDGs.