Combining point cloud slicing and terrestrial laser scanning data to retrieve an effective leaf area index

Abstract

Leaf Area Index (LAI) is one of the most important forest canopy biophysical parameters and plays a key role in many forest biophysical processes and energy and matter exchange among forest-soil-atmosphere. Terrestrial laser scanning (TLS)-based Point Cloud Data (PCD) implicitly contain three-dimensional forest structural information. PCD can be used to quantitatively characterize the structural information of forests from a three-dimensional perspective and provides a high spatial resolution. Consequently, PCD provides highly detailed information about forests and enable the retrieval of forest parameters. In this study, we retrieved the canopy effective leaf area index (LAIe) by using TLS and validated the TLS-based results through field-based measurements. Three-dimensional PCD were acquired with a single- or multiple-location experimental setup in all forest plots. All PCD were clipped into a circular plot with a radius of 30 m. An algorithm called " radial hemispherical point cloud slicing (RHPCS)" was utilized to process the original PCD to retrieve the " Angular Gap Fraction (AGF)" and extinction coefficients by dividing the hemispherical volume into many " trapezoid voxels." A " gap" voxel is an empty or a foliage voxel depending on the number of points within it. By reconstructing the normal vector of a point within a certain neighbor region, we computed the leaf mean inclination angle and obtained the canopy extinction coefficient by using Campbell's algorithm. We calculated the LAIe of each forest plot on the basis of Beer's law by dividing a hemisphere PCD into 9° intervals in the inclination direction. Ten annulus hemispherical regions were obtained. Each region had an azimuthal angle ranging from 0° to 360° and a 9° inclination angle range. Then, we computed the AGF, extinction coefficient, and LAIe of each hemispherical region. We obtain the LAIe of the entire plot by computing the means of the LAIe of 10 hemispherical regions. Results showed that the TLS-based LAIe captured 70.84% (N=9, p<0.01) and 74.09% (N=14, p<0.01) of the variations in LAI-2200 and Digital Hemispherical Plotography (DHP), respectively. We conducted a sensitivity analysis of laser beam angle in AGF estimation and discovered that the AGF of RHPCS should be determined based on a user-predefined laser sampling space and is affected by the characteristic size of foliage elements of a forest canopy. We also analyzed the effect of the number of neighboring points on LAIe calculation. The optimal number of neighboring points was 12 in this study, and we suggest that the number of neighboring points be based on leaf curvature when reconstructing a normal vector. Comparison of TLS-based LAIe with the LAIe results of LAI-2200 and DHP showed that calculating the LAIe of forest plots by combining TLS-based PCD with the RHPCS algorithm is feasible and suitable for forests with different densities and ages. The TLS-based LAIe estimation approach presents considerable importance in quantitatively characterizing forest canopy structural parameters from a 3D perspective.