UAV Imagery to support individual tree management and monitoring

Abstract

Unoccupied Aerial Vehicles (UAVs) have gained prominence in remote sensing applications, a.o. in individual tree detection (ITD) and management. This study explores the potential of UAV imagery in two key applications: assessing tree health using multispectral data on the one hand and evaluating the impact of different Structure-from-Motion (SfM) software packages on ITD on the other hand. In the first case study, multispectral UAV imagery was collected from an urban park in Belgium. Normalized Difference Red Edge Index (NDRE), oftenly used as a proxy of crop health, was found to be related to traditional visual tree assessments (VTAs), which highlights the potential of UAVs to complement ground surveys by offering a 'top view' perspective. While limitations exist in capturing within-canopy information, such as branch structure, multispectral UAV imagery can still be useful in detecting early signs of stress, presenting an objective supplement to subjective expert-driven VTAs. In the second case study, the impact of SfM software packages on ITD was investigated using existing imagery. Pix4DMapper, Agisoft Metashape and WebODM were employed to generate 3D point clouds for individual tree delineation. Pix4DMapper and Agisoft Metashape outperformed WebODM in terms of correctly classified trees, demonstrating higher recall, precision, and F-scores. WebODM exhibited a higher fraction of false negatives and false positives, and detected smaller trees on average. The findings emphasize the importance of SfM software selection in optimizing the accuracy of ITD from UAV-derived data. In conclusion, this research underscores the potential of UAVs in enhancing tree inventories, both in an urban or forest related context. The combination of multispectral imagery for tree health assessment and careful consideration of SfM software choices for ITD can contribute valuable insights to (urban) forestry management.