Estimation of the leaf area density distribution of individual trees using high-resolution and multi-return airborne LiDAR data

TL;DR

This paper presents a novel airborne LiDAR-based method for accurately estimating the leaf area density distribution of individual trees by utilizing last and intermediate laser pulses, improving inner crown foliage mapping.

Contribution

The study introduces a new LAD estimation method that leverages additional LiDAR pulses, enhancing accuracy over previous contact frequency-based approaches.

Findings

Using last and intermediate pulses improves LAD estimation accuracy.

Optimal voxel size for LAD distribution is 1m x 1m x 0.5m.

Method validated with terrestrial LiDAR data on Z. serrata.

Abstract

The three-dimensional structures of individual trees are important pieces of information necessary to understand the effect of trees on urban environments. In this study, we demonstrate a method for estimating the leaf area density (LAD) distribution of individual trees using high-resolution airborne LiDAR. This method improves upon the previously proposed method, which calculates LAD based on the contact frequency between the laser beams and leaves by tracing the paths of the laser beams. The proposed method in this study exploits the last and intermediate pulses in addition to the first and single pulses to capture the foliage distribution in the inner part of the crown. Each laser beam is traced from a point derived by the last pulse to the point derived by the first or intermediate pulse that is recorded immediately before the last pulse. The laser beam interceptions and intersections can thus be accurately reproduced while considering the last and intermediate pulses. We verify the estimation accuracy of the three-dimensional LAD distribution using terrestrial LiDAR data from a single tree (Z. serrata). The appropriate voxel size for representing the LAD distribution from the airborne LiDAR is first determined by comparing the distribution of voxels containing one or more airborne LiDAR points with that containing one or more terrestrial LiDAR points. The estimated LAD distribution with a voxel size of 1 m by 1 m by 0.5 m is subsequently compared to the terrestrial LiDAR-derived LAD distribution. When only the first and single pulses are used, the LAD is overestimated and underestimated in the upper and lower part of the crown, respectively. We confirmed that using the last and intermediate pulses improves the estimation accuracy of the entire crown area.