Urban trees play a significant role in improving urban environment for example removing dust, reducing urban heat islands, damping peak flow, etc. However, the magnitude of welfare provided by trees for urban environment is affected by tree species. Spatially explicit information of species is increasingly needed for understanding the ecosystem services values of urban trees and developing strategies for improving urban environment. And the identification of species accordingly serves as the basic and essential step for sustainable development of urban environment. The state of the art terrestrial level light detection and ranging (LiDAR)—handheld laser scanning (HLS) shows the potential to obtain species information effectively and efficiently.

The structural parameters derived from 3D models or point clouds are commonly utilized to classify tree species from terrestrial LiDAR (TLS). Nevertheless, the performance of combinations of structural parameters derived from different sources is rarely exploited at present. To bridge the gap, the first study of this project is to conduct species classification from HLS data using combination of 3D-model-derived point-cloud-derived parameters. Furthermore, with the rapid development of computer vision techniques, plenty of 3D deep learning (DL) algorithms have been proposed for point clouds classification. Considering little research classifies tree species using 3D DL method, the second study of this project expects to apply 3D DL algorithms to the field of species classification.

Stonewall trees (SWTs) is a uncommon ecological and landscape scene where trees grow on old masonry walls in the urban. Hong Kong (HK) government is paying much attention on the preservation of SWTs because they connect nature and culture and are an illustration of synergy between nature and culture. HLS is a great way to study the structure of SWTs without destroys. Therefore, the third study intends to measure basic parameters of SWTs with HLS, providing more references for later studies.