Application of LiDAR Data to Derive Key Forest Structure Metrics

Details

Supervisors

Defourny, Pierre ; Kmoch Alexander

Faculty

Faculté des bioingénieurs

Degree label

Master [120] en sciences agronomiques et industries du vivant, à finalité spécialisée: Geo-Information Science and Earth Observation for Environmental Modelling and Management

Abstract

enThis study investigates the accuracy of airborne LiDAR in deriving individual tree-level structural metrics relevant to forest management in temperate deciduous forests. Five key metrics—tree height, diameter at breast height (DBH), crown base height (CBH), crown diameter, and aboveground biomass (AGB)—were assessed across four structurally diverse parcels in the Bois de Lauzelle (Belgium), using both LiDAR and field data. A novel individual tree segmentation (ITS) workflow was developed by extending an existing algorithm (PTrees). The method incorporates a Random Forest model to identify segmentation errors and applies an iterative correction process to address over- and undersegmentation. The workflow was benchmarked against a standard CHM-based method and evaluated using a custom in situ matching algorithm designed to handle spatial uncertainty in field data. The novel segmentation approach outperformed the benchmark in all parcels, reducing both omission and commission errors. Structural metrics were extracted from the corrected segments and validated with in situ data. Results showed promising accuracy for tree height (RMSE = 4.23 m, R² = 0.58), though tall trees tended to be underestimated. Crown diameter showed low correlation (R² = 0.11) but aligned well with field distributions. DBH estimates had mixed results (RMSE = 20.66 cm, R² = 0.24), with a consistent underestimation trend. AGB, estimated via allometric equations and regression models, reached similar performance (R² = 0.21–0.22), with slight underestimation and varying accuracy across forest conditions. The only metric showing poor estimation was CBH (RMSE = 9.08 m, R² = 0.01), largely due to overestimation and a lack of low values. The findings confirm that moderate-density airborne LiDAR, combined with calibrated segmentation and validation, can produce usable structural metrics. The method offers a scalable and transferable approach to forest inventory and monitoring, with strong potential to support data-driven forest management in structurally diverse temperate forests.