Experiments with height-diameter models for urban trees using drone LiDAR

Abstract

Tree height-diameter (H-D) models were originally developed to address the difficulty associated with measuring tree height in the field. However, advances in remote sensing, particularly in light detection and ranging (LiDAR), now enable tree height and canopy-width measurements across large spatial extents. Since the accuracy of these measurements is not well understood, we collected LiDAR data using remotely piloted aircraft and evaluated the effects of point cloud density (7-27,800 pts/m2) on height and canopy width measurements. The combination of aerial data and inverted H-D models can be leveraged to estimate diameter at breast height (DBH) and subsequently biomass and carbon content for urban trees. However, existing H-D models are derived from natural forests and their accuracy for urban applications is unknown. To overcome these challenges, we identified twenty frequently used H-D models from the literature and inverted them to predict DBH from canopy height for urban trees. We also constructed new local H-D models as well as models that included canopy width (i.e., H-W-D models) and compared all results against in-situ DBH measurements. Results demonstrated that a minimum point density of 100 pts/m² is required for reliable height and canopy width measurements in urban environments. Although forest-derived H-D models can be recalibrated for application to urban trees, their accuracy was moderate (mean absolute error of 4.9cm DBH). Incorporating canopy width improved model fit (increase of R2 by 0.1) and reduced DBH mean absolute error to ~3.9cm. Results may facilitate the generation and maintenance of urban tree inventories.