Influence of geolocation misalignment on airborne laser scanning-based forest biomass model calibration in structurally complex mixed forests

Abstract

Accurate alignment between forest inventory plots and corresponding airborne laser scanning (ALS) point clouds is critical for reliable area-based model calibration and biomass estimation; however, geolocation misalignment remains a significant challenge in structurally complex mixed forests. In this study, we investigated the extent to which plot-level misalignment influences aboveground biomass (AGB) model calibration and wall-to-wall AGB mapping in a mixed-forest environment. We collected forest inventory data from 56 fixed-area plots (40 for model calibration and 16 for independent validation) in the Howland Research Forest, Maine, USA, and used a professional-grade Trimble GNSS receiver to measure plot center locations. To accurately align inventory plots with the ALS point cloud, a rigid-body misalignment improvement method was developed to refine plot geolocations based on spatial correspondence between field-measured tree stem positions and ALS-derived tree crown centroids. Random Forest models were then calibrated using the improved geolocations and controlled misalignment scenarios simulated at levels of 1, 3, 5, 10, and 15 m, and these calibrated models were applied to estimate AGB for the entire study region. Results suggest that model calibration remained relatively stable at misalignment distances of 5 m or less (R2 ≥ 0.50, RMSE ≤ 39 Mg/ha). However, model accuracy and map agreement declined sharply when misalignment exceeded 5 m. Although regional mean AGB appeared consistent across scenarios, spatial patterns varied substantially and by more than 6% when misalignment exceeded 5 m.