Abstract
With the significant progress of urbanization, cities and towns are suffering from air pollution, heat island effects, and other environmental problems. Urban vegetation, especially trees, plays a significant role in solving these ecological problems. To maximize services provided by vegetation, urban tree species should be properly selected and optimally arranged. Therefore, accurate classification of tree species in urban environments has become a major issue. In this paper, we reviewed the potential of light detection and ranging (LiDAR) data to improve the accuracy of urban tree species classification. In detail, we reviewed the studies using LiDAR data in urban tree species mapping, especially studies where LiDAR data was fused with optical imagery, through classification accuracy comparison, general workflow extraction, and discussion and summarizing of the specific contribution of LiDAR. It is concluded that combining LiDAR data in urban tree species identification could achieve better classification accuracy than using either dataset individually, and that such improvements are mainly due to finer segmentation, shadowing effect reduction, and refinement of classification rules based on LiDAR. Furthermore, some suggestions are given to improve the classification accuracy on a finer and larger species level, while also aiming to maintain classification costs.
Highlights
Rapid urbanization has become one of the most characteristic phenomena of modern times worldwide and has led to important social, economic and environmental consequences [1]
The limitation of optical remote sensing imagery to accurately map urban tree species is attributable to three major reasons: (1) the various surroundings of urban trees create a complicated background, increasing the complexity of classification; (2) overlapping and shadowing effects restrict the segmentation of individual trees or crowns, especially for small-size species; and, (3) the Hughes phenomenon [52], or the curse of dimensionality, that is, given a fixed sample size, the identification accuracy first increases declines with the increase in spectral resolution due to increasing within-species spectral variation [53]
Given the limitations inherent in optical remote sensing imagery and the difficulties posed by unique urban environments, light detection and ranging (LiDAR) has been valued for providing important complementary types of information, such as elevation data and structural features, that have the potential to improve tree species classification accuracy [63]
Summary
Rapid urbanization has become one of the most characteristic phenomena of modern times worldwide and has led to important social, economic and environmental consequences [1]. Urban vegetation, trees, plays important roles in adsorbing air pollutants such as ozone and sulfur dioxide, mitigating carbon dioxide, lowering urban temperature, and dampening peak flows [11,12]. To maximize these benefits, suitable urban tree species should be carefully selected and optimally arranged. With the rapid development of remote sensing technology and the appearance of classification algorithms for computers, dominant species identification at plot level and urban tree species composition prediction have been extensively undertaken based on optical imagery with excellent accuracy [16]. We provide some future considerations for improving the classification accuracies in larger species groups or at finer species levels, hopefully to provide a reference for more accurate and finer urban tree species classification in China and other cities worldwide, to promote urban forest management and urban ecological assessments
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
