Résumé
Upcoming satellite missions targeting the estimation of forest Above-Ground Biomass require an expansion of calibration and validation capabilities. Terrestrial Laser Scanning (TLS) has been demonstrated to be an unbiased estimation tool for single tree wood volume and AGB, especially in large trees. However, TLS field data acquisition is labour-intense and time consuming, and might be overcome by Unmanned Aerial Vehicle Laser Scanning (UAV-LS). In this context, the aim of this study was to explore the potential of individual tree metrics derived from automatically segmented UAV-LS point clouds to estimate tree wood volume across a range of forest sites with varying structural complexity. Four sites were involved, one temperate mixed, two tropical wet and one savanna forest site. Each was surveyed with both TLS (RIEGL VZ-400) and UAV-LS (RIEGL VUX-1UAV). Based on TLS point clouds, reference trees were segmented and wood volume was estimated via Quantitative Structural Modelling (QSM). The UAV-LS point clouds were automatically segmented and trees corresponding to the TLS reference trees were identified. Then, a range of individual tree metrics was derived from the UAV-LS point clouds and compared with TLS reference wood volume. Results showed asymptotic behaviour of metrics tree height and crown diameter (R2 0.16 to 0.78), while crown area, tree volume and novel graph-based metrics showed more linear relationships with volume (R2 0.22 to 0.88). Overall, the available metrics offer promising opportunities for wood volume modelling and future biomass estimation. This will allow the estimation forest biomass across hectare-scales for efficient satellite mission calibration and validation.