Improving urban tree species classification with lidar-derived metrics

Urban tree species mapping provides valuable insights into the green infrastructure management of cities. However, information on the spatial distribution of tree species in urban areas is usually acquired with costly procedures such as field surveys. Remote sensing combined with field data provides an efficient way to obtain spatially explicit information on tree species distribution over broad spatial extents. In this study, we investigate the utility of light detection and ranging (LiDAR) metrics to improve tree species classification in a highly diverse tropical urban setting. LiDAR metrics were estimated using a statistical approach that retrieved surface normals. Moreover, we explore the use of LiDAR reflectivity intensity and canopy height to discriminate among species. The results show that intensity and canopy height improve the classification accuracy, while the use of surface normals reduces it. However, more research is needed to evaluate the utility of surface normals since the species have highly variable patterns, particularly in the nz direction.