Derivation, Validation, and Sensitivity Analysis of Terrestrial Laser Scanning-Based Leaf Area Index

Abstract

. Terrestrial laser scanning (TLS) is a promising tool, used to retrieve leaf area index (LAI). However, the accuracy of LAI estimations based on TLS is still difficult to validate, because high-fidelity destructive measurements of leaf area are lacking. A comprehensive analysis of the sensitivity of TLS-based LAI estimates against various influencing factors (e.g., noise points, woody points, and voxel size) has yet to be reported. We acquired the true LAI by destructively measuring all leaves of 17 magnolia trees. We also improved a voxel-based method to estimate the LAI from the TLS data. We further assessed the sensitivity of LAI estimates against denoising, separation of woody points from foliage points, and voxel size. Our results showed that TLS-based LAI estimations were significantly related to the destructively sampled LAI (R2 = 0.832, RMSE = 0.693). Denoising improved the TLS-based LAI accuracy with a decrease of 0.415 in RMSE. Conversely, wood-leaf separation showed little effect on the accuracy of LAI estimation. The voxel size was an important parameter affecting the accuracy of TLS-based LAI, and our new method for determining voxel size (R2 = 0.832) proved to be more effective than the existing 2 methods (R2 = 0.661 and 0.581).